ASA Connect

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value. 
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk. 

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written. 

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship. 

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"? 

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

I think point 2 is more important than point 1. There no longer is a sharp boundary at p=0.05, and personally I would not object strongly to the term "trend" in such cases. I think the estimates, together with their confidence intervals, are important in such cases. If the estimate and one confidence bound are interpretable as a substantial relationship, this means that such a relationship is compatible with the data. Of course, the other confidence bound is in the opposite direction in such cases, meaning that a zero or opposite relationship is also compatible with the data, but this bound may (or may not) be interpretable as no substantial relationship. As to point 2, it is flatly wrong to state odds ratios as risk ratios, except that it may be approximately ok for very small odds ratios. I would have retracted co-authorship in such cases, yes. I have done that in some cases, mostly because of disagreement with the statistical methodology and less due to issues of presentation. But I have not retracted my cooperation with the article in question, and both professional and possibly other relationships have survived. Concerning odds ratios the interpretation is fairly clear-cut for a dichotomous independent variable. Not so for a continuous independent variable. Then, an odds ratio of say 10000 may be interpretable as tiny, and an odds ratio of say 0.0001 may be interpretable as huge. Usually, the odds ratio for a continuous independent variable should be stated for an interpretable difference of neither tiny nor huge size in the independent variable in question. This is all to often overlooked.

------------------------------
Tore Wentzel-Larsen
Researcher
Norwegian Centre for Violence and Traumatic Stress Studies,
Regional Center for Child and Adolescent Mental Health, Eastern and Southern Norway]
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

1. I am not as bothered as some people by calling 0.06 or 0.07 a "trend". It isn't a trend over time -- it's a commonly used way of saying "almost but not quite significant".  Since many statisticians object to the arbitrariness of 0.05 anyway, I can live with that, as long as the actual p values (or the range) is clearly indicated so readers can judge for themselves  HOWEVER, if the "trends" were described in the abstract or discussion as things "found" without the qualification of non-significance, I'd be much less happy! 

2. So, were the percentages small (such that the odds ratio would approximate the increase in risk?) Did you point out the math (that an increase from 50-50 to 90% is not even a doubling of risk but would be an odds ratio of 9?)  I don't think I would remove my authorship for this one either, as long as the odds ratio was clearly labeled as such -- I've seem papers where they call it a relative risk or risk ratio, Also, remember that the percentages don't have to be small in the samples, only in the population, So in a case-control study that has 50% with breast cancer and 50% without it, the odds ratios do approximate relative risks, because in the population sampled from breast cancer is typically less than 10% of subjects.

Ed

------------------------------
Edward Gracely
Drexel University
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

It is certainly ethical and appropriate to withdraw your name, if you feel that the paper will make claims that are not validly supported by the data.  As others mentioned here, this issue is especially salient for the claims being made about impacts on risk.  (Re the p-values, if the effect sizes and confidence intervals look suggestive on their own, which is fair for the paper to report, it's a bit misleading to imply that the largish p-value adds meaningful additional supporting evidence; but that's maybe a gray area. )

I empathize with you for the bullying you're facing by the lead author.   I was Chair of a university Research Ethics Board for several years; and one  definite stress for the role was when hearing from a researcher (and possibly their dean) how much more prestigious their research and background is than others', and how long they've followed a certain research procedure that involves human subjects--for which either the guidelines have become more sensitive and been revised, or perhaps the procedure should have been questioned in an earlier ethical review!

------------------------------
William (Bill) Goodman
Professor (retired) and Adjunct Professor, Faculty of Business and Information Technology
University of Ontario Institute of Technology
------------------------------

Original Message:
Sent: 02-08-2019 07:27
From: Edward Gracely
Subject: Authorship concern

1. I am not as bothered as some people by calling 0.06 or 0.07 a "trend". It isn't a trend over time -- it's a commonly used way of saying "almost but not quite significant".  Since many statisticians object to the arbitrariness of 0.05 anyway, I can live with that, as long as the actual p values (or the range) is clearly indicated so readers can judge for themselves  HOWEVER, if the "trends" were described in the abstract or discussion as things "found" without the qualification of non-significance, I'd be much less happy!

2. So, were the percentages small (such that the odds ratio would approximate the increase in risk?) Did you point out the math (that an increase from 50-50 to 90% is not even a doubling of risk but would be an odds ratio of 9?)  I don't think I would remove my authorship for this one either, as long as the odds ratio was clearly labeled as such -- I've seem papers where they call it a relative risk or risk ratio, Also, remember that the percentages don't have to be small in the samples, only in the population, So in a case-control study that has 50% with breast cancer and 50% without it, the odds ratios do approximate relative risks, because in the population sampled from breast cancer is typically less than 10% of subjects.

Ed

------------------------------
Edward Gracely
Drexel University

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Hello Jackie,
First, I think one should always feel justified in removing their name if they feel they cannot take ownership of the manuscript and its contents, at least to a reasonable degree in the aspects of it to which they contributed. 

And I do think describing odds ratios as risk is problematic. I was less concerned (although others will undoubtedly strongly disagree with me) about the "trend towards an increase/decrease" comment as I think the readership of medical journals know exactly what that means. I fully understand the concerns about this wording, so - while I would certainly make edits in the paper to change, it's not something for which I would pull my name off of the paper.

I think the issue to consider as a PhD candidate is the balance between providing input on one hand, and demanding changes on the other. On one hand, statisticians need to be the gate keeper on using appropriate methods and correct interpretation, but on the other hand, manuscript writing is a multidisciplinary team effort, and - in my opinion - there does need to be a balance to maximize collaborative impact. I can't say which is right or wrong for you in this situation.

Further, what I cannot tell from your description is if 1) this is a well written paper with rigorous analyses which will contribute positively to the literature but with a couple of places where it could have been worded better, or 2) it's a poorly conducted study which will contribute negatively to the literature. I've been involved with both and have pulled my name off of papers on occasion, but if I demanded the exact wording I wanted every time, I think I would be doing more harm than good. Just my opinion though - I think others take a different stance.

So, you need to judge that for yourself. Positioning yourself to work with the right people is also important. I do not want to judge this investigator, but I usually find that arguing points based on number of past publications, rather than merits of the argument, is a bad sign. I'll admit that I did that once with a student who I was incredibly irritated with for various reasons, but I think it is usually a sign of being unable to argue the merits of the case.

Finally, while I think checking with a more experienced statistician is a good idea, I would do yourself before responding, not offer to do it later. If you feel you need to confirm an approach and have not, you probably should not raise the issue in the first place, or raise it as a suggestion with the idea that you will get back to them more definitively later. If you are confident in the comment, no need to offer to check with others. You can still do it yourself.

Hope that's helpful advice. 
Best, Doug

------------------------------
Douglas Landsittel
Professor of Biomedical Informatics, Biostatistics, and Clinical and Translational Science
University of Pittsburgh
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Jackie,

Excellent question, and one that I suspect many statisticians in medicine (and other fields, too) will relate to.  Both issues that you raise are pet peeves of mine as well and are extremely common misuses of statistical language in the medical literature (with "trend towards significance" offending me more , but YMMV on that!).  As for the OR - it's literally in the name of the measurement, the "odds ratio" - but medical journals very frequently cite OR=4 as "4-fold increase in risk" without appreciating the difference.

The authoritative "I've written XXX papers and been first/senior author on XXX papers; I know what I'm doing; leave the medical writing to me" is exasperating, but (IMO) fairly common among clinical investigators who perceive themselves as the true experts, and view statisticians as service personnel or consultants rather than experts in their own right.  We have all seen or will see this person at some point in our careers.   

The difficult question you ask is whether these issues (which are mildly exasperating, but ultimately not on the level of misconduct or fraud; just poor uses of terminology) are valid reasons to ask to be removed from authorship.  As much as I would love to give a ramrod-straight answer that "you are responsible for the contents of all papers and should never let your name appear if you are in disagreement with anything in the paper" - I think the real-world answer is "it depends" - on where you are in your career, how much you *need* the paper(s) to demonstrate productivity, and whether you are dependent on this person for funding or future work.  

Further complicating this, of course, is the personal friendship that you have with this collaborator.

I will disclose that I have seen co-authors use similar language (many times) and, even after my recommendation to remove or modify, retained the language against my advice.  I have very seldom asked to be removed from the paper for something like this unless they have made other issues so severe that I think the entire message of the paper is compromised (as much as we roll our eyes at "trending towards significance" - I don't think that completely changes the interpretation of the paper).  Others may feel differently, but I did/do not view these as such grievous misstatements that it was worth the conflict (and especially for students and/or junior faculty, you may not in a position to rock the boat).  My typical answer would be to speak with your center director, advisor, or a friendly professor to see what they think or if they might be able to talk some sense into your collaborator, but that might not be as easy as it sounds.

Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"? 

This is the great question facing statisticians who are inherently viewed as "collaborators" rather than experts by many fields.  The pull of inertia from their prior experience or other things they have seen published is so strong.  I think the best you can do in such cases is to respectfully explain your position, cite references or other resources when appropriate (this brings up an important point - be aware of things like EQUATOR network, CONSORT guidelines, etc - these can be your friend, when you show someone a source of "authority" they may respond better than simply telling them yourself).  Some folks have suggested the clever "I was just reading about this the other day..." to make it seem less like you're haughtily telling them why they're wrong - something I admittedly can struggle with.

Thanks for sharing your experience, and hopefully this was at least cathartic, even if I cannot provide much concrete advice...

------------------------------
Andrew D. Althouse, PhD
Assistant Professor of Medicine
Center for Research on Health Care Data Center (CRHC-DC)
Center for Clinical Trials & Data Coordination (CCDC)
University of Pittsburgh School of Medicine
200 Meyran Avenue, Suite 300
Pittsburgh, PA 15213
Email: ada62@pitt.edu
Twitter: @ADAlthousePhD
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

I don't have much to add here that has not been covered, but here are two thoughts.

First, people outside of statistics tend to be a little 'loose' in describing results, using ways that we were told NOT to do it that aren't technically correct. Conflating risk and odds, and describing p-values as 'trending toward significance' are very common in other disciplines.  You have to do what's right for you in terms of authorship, but these aren't show stoppers for me if the research of high quality.  I try to get people to focus on the estimates and their precision vs. being mesmerized by p-values. 

My bigger concern for you is in mending the relationship.  It seems that you and your friend both reacted a bit strongly based on what you describe -  As I don't know either of you, I have no advice here.  However, a face to face discussion in the future may diffuse the contest about who knows best that seemed to occur over e-mail.

Good luck.

Rachael

------------------------------
Rachael DiSantostefano
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

You are correct on both issues although odds ratios are related to relative risk for small prevalence, not risk per se. I think not seeking authorship on the paper is appropriate since it would not reflect well on your skills. You are not in a commanding position as a PhD candidate giving advice to a seasoned researcher even though she would benefit from your advice. Even as a full professor, on a couple of occasions I was dropped from research committees when my design/stat recommendations were in conflict with the chair. I suggest putting your recommendations into a memo to the researcher along with your desire not to be among the authors just to document the situation. There is probably no real strategy to suggest for future situations of this type other than get your PhD, get tenure and then you magically gain clout.

------------------------------
Chauncey Dayton
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

You can hope that the reviewer calls her out on both.  Especially as a junior statistician that is one of the only things that can work.  (And it's glorious when you get feedback that echos what you've said).

#1 is a fairly common error that people write -- and hopefully readers understand -- but it's flat out wrong as written.  But agree with Ed's example, and it is the one I always use too to illustrate RR vs. OR.  50% to 90% is a RR = 0.8 and OR = 9 (!!!)  

Regarding p-values, good luck. You're going to have that fight a lot more.  Authors want to salvage their work and they do that via spin -- that's what she's doing and that probably half the other manuscripts in the issue of the journal are doing too.    There is a great Twitter account "I_need_a_p_hack" that quotes phrases people write for their P>0.05 results. It's quite hysterical.

(Plus given Ho is true (and she started with that assumption)  the probability = 0.07 is EXACTLY THE SAME as p=0.9.  ​The VALUE of a p-value tells is little -- but that's a whole different argument).

The bigger lesson here is that good statisticians are sufficiently sought after that you can be judicious in your collaborations.  You can choose to work only with collaborators who will listen and take your advice, and who will compromise.  (Sometimes we have to work with the not so good ones to learn who they are).  It gets easier to be choosy as you get more experience -- but choosing wisely is a talent worth developing and will make your career very fulfilling.

(And good job not saying "You have 145 papers because apparently you don't take the time to write carefully and accurately.")



------------------------------
Jason Connor
ConfluenceStat, LLC
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

What my former Cleveland Clinic colleague and Saturday golf buddy Jason Connor "failed" to mention is that he wrote a terrific, pithy article that is very relevant to this discussion, an article that statisticians can give to investigators (and, still, some statisticians) who are confirmed p-value-ists. When I was teaching my "unique" two-semester graduate-level course in statistical science, we read and discussed this in the first week or two, and it helped greatly to set the stage for the entire year. 

Here's the PDF: https://s3-us-west-2.amazonaws.com/misc.biblio/Connor04The+value+of+a+p-valueless+paper.pdf

Isn't this a great title?

------------------------------
Ralph O'Brien
Professor of Biostatistics (officially retired; still keenly active)
Case Western Reserve University
http://rfuncs.weebly.com/about-ralph-obrien.html
------------------------------

Original Message:
Sent: 02-08-2019 11:08
From: Jason Connor
Subject: Authorship concern

You can hope that the reviewer calls her out on both.  Especially as a junior statistician that is one of the only things that can work.  (And it's glorious when you get feedback that echos what you've said).

#1 is a fairly common error that people write -- and hopefully readers understand -- but it's flat out wrong as written.  But agree with Ed's example, and it is the one I always use too to illustrate RR vs. OR.  50% to 90% is a RR = 0.8 and OR = 9 (!!!)

Regarding p-values, good luck. You're going to have that fight a lot more.  Authors want to salvage their work and they do that via spin -- that's what she's doing and that probably half the other manuscripts in the issue of the journal are doing too.    There is a great Twitter account "I_need_a_p_hack" that quotes phrases people write for their P>0.05 results. It's quite hysterical.

(Plus given Ho is true (and she started with that assumption)  the probability = 0.07 is EXACTLY THE SAME as p=0.9.  ​The VALUE of a p-value tells is little -- but that's a whole different argument).

The bigger lesson here is that good statisticians are sufficiently sought after that you can be judicious in your collaborations.  You can choose to work only with collaborators who will listen and take your advice, and who will compromise.  (Sometimes we have to work with the not so good ones to learn who they are).  It gets easier to be choosy as you get more experience -- but choosing wisely is a talent worth developing and will make your career very fulfilling.

(And good job not saying "You have 145 papers because apparently you don't take the time to write carefully and accurately.")



------------------------------
Jason Connor
ConfluenceStat, LLC

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

That's an unfortunate situation, but does come up from time to time (especially early in your career).

The issues themselves are rather trivial. Yes, it's incorrect to say "trend towards significance" and to interpret relative odds increase as relative risk increase. You did very well by offering authoritative references and the option of consultation w/ a senior statistician. Both failed.

If this was a consult within an institution (eg, Consulting Center), then you absolutely have to go to your manager/director/chair.

If this was a private consult, under these circumstances, I absolutely take my name off the paper (it has happened to me). Most importantly, I never again work with these investigators. Not because of the 2 specific issues, but because of a bigger problem: their lack of trust in my skills/expertise and lack of respect and appreciation of my work. If they know better, they don't need me, they should do their stats on their own, thank you very much.

Sad and frustrating, but there are better people out there with whom to work.

Constantine

PS Semantics but... a 4.0 OR is not a 4-fold increase in odds. The odds of one group are 4 times the odds of the other, but the increase in odds is only 3-fold (OR-1), ie, the increase/difference in odds is only 3 times the odds of the control/comparator group.

------------------------------
Constantine Daskalakis, ScD
Thomas Jefferson University, Philadelphia, PA
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Jackie,

To answer your two questions in order:

1. Yes, asking to be removed from authorship of the paper was the right thing to do. Allowing your name to be on the paper would be sanctioning poor statistical practice. "That's the way we've always done it" is not a good or ethical reason to do something.

2. This is much more difficult. Once upon a time, I thought that doing my best to give a tactful reply pointing out misunderstandings would be taken seriously, but my experience has been disappointing -- more often than not, the person whose work I was commenting on just brushed off the criticism as, at best,  a minor point. A big part of the problem is that statistical practice is all too often taught in a "This is the way we do it" manner. That's what prompted me to develop a continuing education course after I retired (materials open source at https://web.ma.utexas.edu/users/mks/CommonMistakes2016/commonmistakeshome2016.html). I don't teach it any more, but have passed it on to someone else. 

Also: The website https://mchankins.wordpress.com/2013/04/21/still-not-significant-2/ should help bolster your confidence that your point (1) is sound. However, I doubt that it would convince someone who "has always done it that way".

------------------------------
Martha Smith
University of Texas
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Both wordings seem OK to me.  A clearly alternative to "trend towards an increase/decrease" might be "borderline significance" or "almost significant".  A sentence recommending repeating the experiment to verify or refute the results might be appropriate.  I've used that approach in internal reports but not in publications.

------------------------------
Emil M Friedman, PhD
emilfriedman@gmail.com
http://www.statisticalconsulting.org
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

I think you did it right. 

In my 30-year career, I removed my name from 4 papers, for reasons similar to yours.  In one situation I had worked with a physician for three years, designing a case-control study to identify possible risk factors for unexplained diarrhea in young children.  I participated in the entire study from conception, design, followup during the data collection, did the data analysis and helped  writing the manuscript.  Then, a co-author, whom I had never meet, decided that the proper name of the study was a "prospective case-control".  That definition that exists in an epidemiology book written by very famous epidemiologists, but is very confusing and it really means that you have two groups, one with a disease (the "cases") and one without (the "controls") and you follow them prospectively for some  outcome other the disease itself.  In my view this should be called a comparative study, since you are just comparing two groups prospectively. 

Regardless that definition, what we had designed was truly a case-control (first we observed the cases with diarrhea, then we found a control matched by age, sex and zip code of the case, and all the variables were collected retrospectively).  We went back and forth on the discussion for six months  about the proper name, and I finally decided to remove my name from the paper.  The group invited  one of the book authors to be in the paper, and later on the paper was published with "case-control" only (not prospective...).  I got no apologies or recognition.  But I would do it again, because there is a point where I cannot go against my principles.  If I always let the first author interpret the results at their will and without regard to our expertise, then we are saying that what we do has no value and our trade is void of meaning. 

Sometimes, when there is no harm, I compromise a little.  In fact, I felt for the "trend" interpretation a couple of times years ago, but no more!  I congratulate you on following your principles.

------------------------------
Marcia Ciol
Research Professor
University of Washington
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Jaclyn,

At first I was going to say how sorry I was that you had to go through this experience, BUT, it is actually a good thing that you went through this experience because this is not the last time you will be in such a situation.  It is encouraging that you are soliciting feedback so that you can learn and grow.

On your first concern, I would not have said "trend towards", but I also would have not said, "non-significant".  I do not think that either is incorrect, but to me, the statements don't mean as much as saying, "there is some evidence that there is an increase/decrease".  The best saying depends on the subject matter and your philosophy.  I cannot comment on the Odds ratio because I do not know enough, but just like the word "significance", the word "risk" may have another meaning depending on the subject matter.

The important thing is your decision to take your name off of the paper.  I completely agree that if you were not comfortable with your name on the paper that you should have it removed.  I like that you suggested a consult with another statistician.  However, it's not always what you say, but how you say it.

"With my limited experience, I am not sure that this is the best wording.  Could we have a more experienced statistician than myself meet with you?"  This approach is much less challenging.  Even saying experienced statistician is much less challenging than PhD Statistician, as if a PhD Statistician trumps anybody else's pedigree.  On pulling off your name, "I am so happy that I was able to work on this project and the results are encouraging.  However, I am still unsure about the wording of the conclusions, which may be perfectly fine.  Perhaps it would be best if I were to be listed as a contributor to the analysis and not an author".

Would I continue to work with this person?  Absolutely.  It is better if they get advice from you than no advice at all.  It is in my experience that if you do good work, listen well and don't have a hidden agenda, that over time your clients will trust your judgement.

Good luck,
Phil

------------------------------
Philip R. Scinto
Senior Fellow
The Lubrizol Corporation
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Hi Jackie,

Great questions. A few comments, (1) first author, senior author, fifth author, etc. they are all co-authors (sometimes people forget that) (2) the statement "Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk," is wrong not only to refer to risks but also that an odds ratio of 4.0 is a 3-fold increase in the odds of the outcome, and an odds ratio of 5.0 is a 4-fold increase in the odds of the outcome, etc.

To answer your questions, in my opinion, mistakes of this magnitude are not cause to remove yourself as an author but it is right to voice your concern. I once saw a letter to the editor written critiquing a paper on which the letter to writer was a co-author :) . That said, this is atypical and your concerns, to me, do not merit such an approach. I hope this helps.

------------------------------
Nelson Pace, PhD, SM
Scientist
Exponent, Inc.
https://www.exponent.com/professionals/p/pace-nelson
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

A sincere thank you to all who have posted here and messaged me individually. I'm a bit more seasoned than my PhD candidacy status implies (I worked as a masters level biostatistician for 11 years prior to returning to school), but have had situations like this come up throughout my career and expect that this will continue, so I've found the responses to be tremendously helpful. Thank you once again and I hope you all have a wonderful weekend.

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

​When I was young, I was in a situation very similar to yours (including the first author's being a friend). The statistical issues, like yours, weren't large but, similar to the ones you list, clearly wrong. So when the first author refused to make the changes,  I asked to be removed as an author. The other authors. who all lived in a publish-or-perish world, were shocked. They couldn't believe that someone - especially early in her career - would remove her name from authorship. That made them realized the error must have been important and changed the paper.

Hold your ground - one paper more or less won't affect your career, but signing on to something statistically incorrect is intellectually uncomfortable.

A word about "trend" - somehow the scientific world has decided that a p-value of 0.05 is "statistically significant" and one of 0.10 is "a trend" - I have no idea where that convention came from, but we statisticians should put an end to it. Your language is much more scientifically sensible.

------------------------------
Janet Wittes
Statistics Collaborative, Inc.
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Addressing only the second question, in my view you don't need any advice: you handled it perfectly. Specifically,

• You had a question about practice, so you researched it (your recommendations were based on "based upon previous posts on the ASA message boards" 
• You stated - in your view and to the best of your professional ability - what is sound statistical practice
• Your statistical recommendations were refused by the lead author, but not for statistical reasons ("I know better than you" is not a mathematical argument) - but you still had still had significant reservations
• Accordingly, you asked your name be dropped from the paper

Well done! You handled this with poise. You don't feel you can support the conclusions as stated, and so asked to have your name removed. In my view, this is exactly as it should be.

Tore Wentzel-Larsen wisely observed "I have not retracted my cooperation with the article in question, and both professional and possibly other relationships have survived" - a point very well taken. 

I would be remiss if I failed to mention there might be consequences. As one example, finding myself in a very similar situation many years ago (different statistical problem but I strongly disagreed wit the lead author on a question of practice), I suddenly found myself unemployed - my contract as a temp was abruptly terminated with no reason given. Despite the short term difficulties, I still regard it as one of  my proudest moments as a scientist. Standing up for good practice really matters, and you have done so under trying circumstances. Brava!

 

------------------------------
David J Corliss, PhD
Director, Peace-Work www.peace-work.org
davidjcorliss@peace-work.org
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Jaclyn,

As you can tell from the number of replies you received, this is a non-trivial concern for many consulting/applied statisticians.  Your concerns about the use of 'trend' and 'risk' are legitimate.  Your concerns about the senior investigator's response are also.  All too often, those of us who have worked primarily as consulting biostatisticians have encountered variations on this theme, as you yourself encountered as a younger statistician, being regarded as mere technicians rather than as trained scientists.  One need only note the recent squabble published in the NY Times between MSK's CEO and Colin Begg to be reminded of this.  Although we are all aware of the downside of taking the position you did - being sent to statistical Siberia or losing one's job - I stand with Janet Wittes on this matter.  As the great philosopher, Janis Joplin, once said, "Don't compromise yourself.  That's all you've got."

------------------------------
Gene S. Fisch, Ph.D.
Adj. Prof., Paul H. Chook Dept of CIS & Statistics,
CUNY/Baruch College
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

This is a related theme, perhaps a little off topic. This article from 1998 https://jamanetwork.com/journals/jama/fullarticle/188182 has had considerable influence, with consistently a lot of citations. In 2018, 52 citations indexed by PubMed. Many researchers feel a need to convert from an odds ratio to a risk ratio, and the article offers a simple solution. In 2003 McNutt et al pointed to problems with this procedure for odds ratios from adjusted logistic regressions, see e. g. a more recent open source discussion in https://academic.oup.com/aje/article/179/8/1034/107994 . This does not seem to have reduced the popularity of the procedure substantially. I have, however, not checked how many of the citations are using the procedure for adjusted odds ratios. Many years ago one of my co-workers among clinical researchers had found the article, I thought the procedure seemed a little to good and made some searches. We did not use the procedure.

------------------------------
Tore Wentzel-Larsen
Researcher
Norwegian Centre for Violence and Traumatic Stress Studies,
Regional Center for Child and Adolescent Mental Health, Eastern and Southern Norway]
------------------------------

Original Message:
Sent: 02-11-2019 12:30
From: Gene Fisch
Subject: Authorship concern

Jaclyn,

As you can tell from the number of replies you received, this is a non-trivial concern for many consulting/applied statisticians.  Your concerns about the use of 'trend' and 'risk' are legitimate.  Your concerns about the senior investigator's response are also.  All too often, those of us who have worked primarily as consulting biostatisticians have encountered variations on this theme, as you yourself encountered as a younger statistician, being regarded as mere technicians rather than as trained scientists.  One need only note the recent squabble published in the NY Times between MSK's CEO and Colin Begg to be reminded of this.  Although we are all aware of the downside of taking the position you did - being sent to statistical Siberia or losing one's job - I stand with Janet Wittes on this matter.  As the great philosopher, Janis Joplin, once said, "Don't compromise yourself.  That's all you've got."

------------------------------
Gene S. Fisch, Ph.D.
Adj. Prof., Paul H. Chook Dept of CIS & Statistics,
CUNY/Baruch College

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Thank you once again to all who took the time to comment here or to message me privately, it is truly appreciated. I wanted to let everyone know that the senior author decided to accept my edits and submitted a manuscript that I now am happy to be a co-author on. We also had a good discussion about our professional relationship and how to handle disagreements, during which many of your comments and suggestions were quite helpful. I sincerely appreciate everyone's guidance, I learned a lot from each post and message. Thank you, all!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------

Original Message:
Sent: 02-07-2019 19:30
From: Jaclyn Szymonifka
Subject: Authorship concern

I recently took on a consulting project and the research team has been sending around a manuscript draft for comments and edits. I had two concerns with the most recent draft that I brought to the team's attention:
(1) Comparisons that yielded p-values between 0.05 and 0.10 were described as a "trend towards an increase/decrease", which based upon previous posts on the ASA message boards, I recommended changing to saying a "non-significant increase/decrease" and explained that with a single comparison, there's no trend and, in fact, a larger sample or another study may yield a statistically significant result, but it also may lead to a higher p-value.
(2) Odds ratios of, say, 4.3 and 5.2 were described as a 4- or 5-fold increase in risk, which I specifically was told this past week in class is only (approximately) correct when the event rate is less than about 10%, but for more frequent events, odds ratios are showing an increase/decrease in odds, but not risk.

The senior author replied that she was the one who wrote the sections I was concerned with, and that they are "fine as is". I responded to her directly and said that these are incorrect and offered to provide her with sources to confirm what I was telling her and also offered to have the two points confirmed by a PhD statistician if she'd like. She said that I was wrong, mentioned that she has 145 publications and has been first or senior author on many of them, and said that I need to leave the medical writing to her as she has much more experience than I do. Realizing that this was a lost cause, I said that she's welcome to do as she pleases but if she keeps the text as is, to please remove my name from the authorship list as I'm uncomfortable stating the results as they were written.

One additional item possibly worth mentioning is that I've worked with this woman on various projects for nearly a decade and we've become close friends, which at times has complicated our working relationship.

Two questions that I was hoping others might be willing to weigh in on:
(1) Were these two issues valid reasons to ask to be removed from the authorship, or are they minor enough to be overlooked?
(2) Generally speaking, can anyone offer some advice on how to handle situations in which a non-statistician is unwilling to accept your statistical "expertise", particularly when their rationale is along the lines of, "This is how I've always done ___" or "Other studies do it this way, so we need to as well"?

Thanks in advance!

------------------------------
Jackie Szymonifka, MA
NYU School of Medicine, Department of Public Health, Division of Biostatistics
PhD candidate
------------------------------