ASA Connect

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

The flashlight idea is a great example!!

There are many dramatic ways to illustrate statistical concepts.  In teaching probability, I use playing cards.  First, I have students calculate various probabilities and then I go around the room and ask students to see if they can pick a certain card from the deck.  Given the low probability of picking the Ace of Spades (.019), it has been very exciting on the two or three occasions when this has happened.  This exercise also gives students a chance to see that actions with low probabilities can occur and actions with high probabilities often do not.  In teaching variability, I slowly rip a sheet of paper into pieces and then toss all the pieces into the air in the middle of the classroom.  I then ask students to find the piece that marks the center of the distribution.  They can then "see" that although we may decide on central tendency that all the other pieces of data scatter away from it and this is variability.  We then go on to figure out how to calculate the average difference.

-------------------------------------------
Donna Chirico
York College The City University of New York
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

I explain correlation using two thermometers, one Fahrenheit, the other Centigrade. What happens to C when F gets higher? (also gets higher) Why does that happen? (Not because F causes C to rise, but because both thermometers measure the same thing--heat). Thus, correlation is am easure of the extent to which two variables measure (some of) the same thing(s).

-------------------------------------------
Francis Dane
Chair
Jefferson College of Health Sciences
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

Elaine,

I like your word-pictures. For me, though, statistical power is more like a magnet's ability to pick up a paper clip. That ability depends not only on the magnetic strength, but on the proximity to the paper clip.  Similarly, statistical power (assuming we are in the Fisherian significance testing paradigm, not the Neyman-Pearsonian hypothesis testing one) depends on both the concentration (peakedness, roughly) of the test statistic's distribution & the deviation, in the population, from the tested hypothesis Ho. Here, the distance corresponds to the deviation from Ho, & the magnet's strength corresponds to the peakedness of the distn.  

Admittedly, this parable is more complicated than flashlights; however, the enhancement serves as a springboard for explaining several important features of testing, such as the confounding in testing of sample size & population effect size, & the possibility of over-powering. For instance, if the sample size is very large, the distn will be very peaked, so that the test might have high power to detect even an inconsequential deviation from Ho. Ideally, we want high power to detect a meaningful deviation from Ho, & low power to detect anything less.  

Also, you might be careful about the word "significance." Significance applies to the sampled data, rather than something in the population; a test statistic is either significant or non-significant. Sometimes people (even statisticians) speak as if whatever is true in the sample is automatically true in the population (in a book I wrote, I call this "direct frequentism"); however, that's not an element of either significance or hypothesis testing the way these systems were set up originally.
-------------------------------------------
Andrew Hartley
Associate Statistical Science Director
PPD, Inc.
-------------------------------------------

Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

I like your flashlight concept Elaine!  I'll  remember that one!

I'm not a golfer myself, but since many of my clients are physicians and golfers, I came up with an analogy that some of them can relate to.  I ask them to imagine going to a new golf course, and I'm their caddie (I held this job for one summer as a kid... never again!)  Some caddies just carry golf clubs and that's it. The better ones know the golf course very well, build relationships with the regular golfers, and make recommendations for which club to use at various times in the game.

Now... I continue...As their caddie, if I don't know the golfer and how consistently she swings (variability),  and if I don't know the lay out of the course either (experimental design) , and if the distance from the to the hole is also a a mystery (effect size),  it would be really difficult for me to recommend an appropriate golf club to use (sample size) to acheive their hole-in-one goal (p<0.05)..

From there, I explain that as their Biostatistician, I need to know the variability of the primary outcomes of interest (how consistently the golfer swings), the experimental design (lay of the land), and the effect magnitude we are looking for (distance to the hole).  With that information, I can usually help them understand the minimum sample size necessary to find what they are looking for (hole in one).

It's a stretch sometimes, but even non-golfers seem to appreciate the analogy.

-------------------------------------------
Robert Ploutz-Snyder, PhD, Pstat(r)
NASA Biostatistician
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

Original Message------

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

Thanks Elaine! I like the way you describe the power and significance.

I wonder if any one could help me to explain a power issue I had.

 

Study:

Physicians want to compare a new drug with a standard drug for relief post-op pain in our hospital (RCT).

 

Sample size calculation:

Sample size 30 and 30 to achieve 80% power to detect a difference of 2 days with pain score >4 between treatment and placebo group assuming standard deviations of 2.7 and with a significant level of 0.05 using a two-sided two sample t-test.

 

Results:

For the final results, there was no statistically significant difference between these two group for the number days with pain scores >4, mean days in treatment group was 1.8±2.1 (SD) and 2.4±3.1 (SD) in placebo group, p=0.85.

 

Physicians insist that the reason we didn't see the significant difference because we didn't enroll enough patients.

 

I want to know what will be the best way to explain the issue to them. Thanks a lot!

 

-------------------------------------------
Yi Lu
Health Services Researcher
-------------------------------------------

Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

One of my favorite concepts on basic stats is the concept of variation. One example I use, especially when the students are american footbal fans, is the decision of a coach on which play a coach might choose in two different situations. The coach has 2 running plays to choose from. One gains an average of 3 yards with a standard deviation of one yard. The second play gains an average of 3 yards with a standard deviation of 10 yards. In a situation where they need to gain two yard for a first down the coach would choose the first play. In a situation where they need to gain 10 yards then the coach should choose the second play.

I might also add a graphic showing the distribution of the yards gained and the how they plays effect the two different situations.

-------------------------------------------
Michael Mout
MIKS
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

Original Message------

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

For power, I use binoculars, explaining that just as the power of a pair of binoculars measures how well they
help you resolve a blob into the actual separate objects, so does the power of statistical test reflect your ability
to distinguish real differences.

For sample size and its relationship to power, there's always the classic joke about the researcher reporting
the results of testing a new drug on rabbits: "A third of the subjects died, a third of the subjects survived, and
the other one ran away." Then you can move on to a more serious discussion of why the results from small sample sizes
may not be informative.

>>Kathy

-------------------------------------------
Katherine Godfrey
-------------------------------------------


Original Message:
Sent: 11-04-2014 13:24
From: Michael Maltz
Subject: What statistical concept can you explain best to others who are not statisticians?

Sample size.

I would tell my class, "Suppose you're making chicken soup foryour family. You cut up the chicken, chop up a few celery stalks and carrots,and add salt. Then after it begins to boil you stir it well and leave it tosimmer for about an hour. But before you serve it you take a spoonful to see ifyou need more salt."

"Now suppose you're making chicken soup for the Eighth Army.You cut up forty dozen chickens, chop up a gazillion celery stalks and carrots,and add a pound or two of salt. Then after it begins to boil you stir it welland leave it to simmer for about an hour. How many gallons of soup do you need todrink to see if you need more salt?"

This gets us into the whole idea of sampling; salt is spreaduniformly throughout the soup, but what about carrots? Some may float to thetop, some may sink to the bottom. How do you make sure that there's enough ofthem? And we go on from there.

--

Michael D. Maltz

maltzmd@gmail.com

http://osu.academia.edu/MichaelMaltz

When I explain the role of power in negative studies to an intro class of public health or med students I often use the dog sniffing drugs analogy.

To show that there are no drugs in your house you want a dog with a good nose, but who doesn't find anything (high power, negative study).

Using a dog with a cold may not find any drugs, but won't be convincing either (low power, negative study).

This is a great thread!

Ed

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


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

I like to use "blowing soap bubbles"
Suppose the variable of interest is time to pop or distance traveled; bubbles that float away without popping can be used to describe censoring. 
What about the attempts that result in no bubbles? 
The number of bubbles for one swipe depends on the soap, wand, and weather conditions (covariates).
The number of bubbles blown depends on the size of the bubbles (correlation).
If there are lots of little bubbles and one large one, it's an outlier; if there's a lot of little ones and a lot of big ones, it's variability.

David
-------------------------------------------
David Bristol
Statistical Consulting Services
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

I teach undergraduate Statistics to science majors
I like to explain the idea of a point estimates and confidence intervals using "Pin the Tail on the Donkey" as an analogy.
It's very unlikely that someone pins the tail "exactly" on the right spot (the "x-mark" on the donkey's bottom). But, if everyone is given a big, "wide" tail, then (1-alpha)% will cover the "x-mark". 


-------------------------------------------
Sara Shain
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

I have sometimes found that referring to the "margin of error" in poltical opinion polls helps peopel understand confidence intervals.   

-------------------------------------------
David Luskin
USDOT-FHWA
-------------------------------------------


Original Message:
Sent: 11-07-2014 18:23
From: Sara Shain
Subject: What statistical concept can you explain best to others who are not statisticians?

I teach undergraduate Statistics to science majors
I like to explain the idea of a point estimates and confidence intervals using "Pin the Tail on the Donkey" as an analogy.
It's very unlikely that someone pins the tail "exactly" on the right spot (the "x-mark" on the donkey's bottom). But, if everyone is given a big, "wide" tail, then (1-alpha)% will cover the "x-mark". 


-------------------------------------------
Sara Shain
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

When I taught probability and statistics as a three week part of a math overview course for liberal arts majors, my students just could not comprehend confidence intervals.  One of the main examples was the margin of error.  I believe the popular conception of MOE is that the parameter lies within the MOE with 100% probability.  I've never seen it explained in the media, just taken for granted.

-------------------------------------------
Charles Coleman
-------------------------------------------


Original Message:
Sent: 11-10-2014 10:42
From: David Luskin
Subject: What statistical concept can you explain best to others who are not statisticians?

I have sometimes found that referring to the "margin of error" in poltical opinion polls helps peopel understand confidence intervals.   

-------------------------------------------
David Luskin
USDOT-FHWA
-------------------------------------------


Original Message:
Sent: 11-07-2014 18:23
From: Sara Shain
Subject: What statistical concept can you explain best to others who are not statisticians?

I teach undergraduate Statistics to science majors
I like to explain the idea of a point estimates and confidence intervals using "Pin the Tail on the Donkey" as an analogy.
It's very unlikely that someone pins the tail "exactly" on the right spot (the "x-mark" on the donkey's bottom). But, if everyone is given a big, "wide" tail, then (1-alpha)% will cover the "x-mark". 


-------------------------------------------
Sara Shain
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

Charles' point is well taken, though not always essential to get the audience to accept one's message. I once presented to a transportation department that had been conducting a survey of parked vehicles for years. Their rule was to sample a fixed percentage of vehicles, so as the years went by the sample size kept growing. I persuaded them that the level of accuracy depends on the size of the sample, not the size of the population, and referred to MOE from politcial opinion polls to convey the concept of statistical accuracy. Some of them may have still had the misconception that Charles would like to dispel, but for my purposes my mini-lecture worked. They bought my argument that the sample size did not need to be increased from the previous survey becasue it was already large enough to ensure a very narrow confidence interval
 

-------------------------------------------
David Luskin
USDOT-FHWA
-------------------------------------------


Original Message:
Sent: 11-11-2014 08:13
From: Charles Coleman
Subject: What statistical concept can you explain best to others who are not statisticians?

When I taught probability and statistics as a three week part of a math overview course for liberal arts majors, my students just could not comprehend confidence intervals.  One of the main examples was the margin of error.  I believe the popular conception of MOE is that the parameter lies within the MOE with 100% probability.  I've never seen it explained in the media, just taken for granted.

-------------------------------------------
Charles Coleman
-------------------------------------------


Original Message:
Sent: 11-10-2014 10:42
From: David Luskin
Subject: What statistical concept can you explain best to others who are not statisticians?

I have sometimes found that referring to the "margin of error" in poltical opinion polls helps peopel understand confidence intervals.   

-------------------------------------------
David Luskin
USDOT-FHWA
-------------------------------------------


Original Message:
Sent: 11-07-2014 18:23
From: Sara Shain
Subject: What statistical concept can you explain best to others who are not statisticians?

I teach undergraduate Statistics to science majors
I like to explain the idea of a point estimates and confidence intervals using "Pin the Tail on the Donkey" as an analogy.
It's very unlikely that someone pins the tail "exactly" on the right spot (the "x-mark" on the donkey's bottom). But, if everyone is given a big, "wide" tail, then (1-alpha)% will cover the "x-mark". 


-------------------------------------------
Sara Shain
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

To supplement (complement?) Mr. Luskin's comment, one could easily give a simple calculation for accurate estimates of the estimates of the margin of sampling error (with the usual 95 percent confidence) for proportions (such as in political POPS) where n << N, which is simply the reciprocal of the square root of the sample size, 1 / sqrt(n).

-------------------------------------------
David Bernklau
(David Bee on Internet)
-------------------------------------------


Original Message:
Sent: 11-10-2014 10:42
From: David Luskin
Subject: What statistical concept can you explain best to others who are not statisticians?

I have sometimes found that referring to the "margin of error" in poltical opinion polls helps peopel understand confidence intervals.   

-------------------------------------------
David Luskin
USDOT-FHWA
-------------------------------------------


Original Message:
Sent: 11-07-2014 18:23
From: Sara Shain
Subject: What statistical concept can you explain best to others who are not statisticians?

I teach undergraduate Statistics to science majors
I like to explain the idea of a point estimates and confidence intervals using "Pin the Tail on the Donkey" as an analogy.
It's very unlikely that someone pins the tail "exactly" on the right spot (the "x-mark" on the donkey's bottom). But, if everyone is given a big, "wide" tail, then (1-alpha)% will cover the "x-mark". 


-------------------------------------------
Sara Shain
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

The easiest ones are the ones they view as good news.  The hardest are the ones they view as bad news.
-------------------------------------------
Emil M Friedman, PhD
emil.friedman@alum.mit.edu (forwards to day job)
emilfrie@alumni.princeton.edu (home)
http://www.statisticalconsulting.org
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

I agree with Emil that "bad news" is harder to explain than "good news." So I sometimes discuss statistical methods in terms of "the bad news" and "the good news" -- see, for example http://www.ma.utexas.edu/users/mks/CommonMistakes2014/WorkshopSlidesDayTwo2014Part1.pdf

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


Original Message:
Sent: 11-13-2014 09:51
From: Emil Friedman
Subject: What statistical concept can you explain best to others who are not statisticians?

The easiest ones are the ones they view as good news.  The hardest are the ones they view as bad news.
-------------------------------------------
Emil M Friedman, PhD
emil.friedman@alum.mit.edu (forwards to day job)
emilfrie@alumni.princeton.edu (home)
http://www.statisticalconsulting.org
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

I was talking about things like "our new product beats the competition" vs "the difference is too small to be of importance" or, even worse, "our new product is not as good as the competition".  Or "the experiment you did doesn't answer your questions".  Etc
-------------------------------------------
Emil M Friedman, PhD
emil.friedman@alum.mit.edu (forwards to day job)
emilfrie@alumni.princeton.edu (home)
http://www.statisticalconsulting.org
-------------------------------------------


Original Message:
Sent: 11-14-2014 15:50
From: Martha Smith
Subject: What statistical concept can you explain best to others who are not statisticians?

I agree with Emil that "bad news" is harder to explain than "good news." So I sometimes discuss statistical methods in terms of "the bad news" and "the good news" -- see, for example http://www.ma.utexas.edu/users/mks/CommonMistakes2014/WorkshopSlidesDayTwo2014Part1.pdf

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


Original Message:
Sent: 11-13-2014 09:51
From: Emil Friedman
Subject: What statistical concept can you explain best to others who are not statisticians?

The easiest ones are the ones they view as good news.  The hardest are the ones they view as bad news.
-------------------------------------------
Emil M Friedman, PhD
emil.friedman@alum.mit.edu (forwards to day job)
emilfrie@alumni.princeton.edu (home)
http://www.statisticalconsulting.org
-------------------------------------------


Original Message:
Sent: 11-03-2014 08:08
From: Elaine Eisenbeisz
Subject: What statistical concept can you explain best to others who are not statisticians?

What statistical concept can you explain best to others who are not statisticians? And, how do you describe it? For instance, I explain the concept of power as being a flashlight, and significance is hiding in the corner of a room. If your flashlight has enough light (power) you will see significance that is truly there, but if your flashlight is too dim (not enough power) you won't see significance even when it is truly there.

-------------------------------------------
Elaine Eisenbeisz
Owner and Principal Statistician
Omega Statistics
-------------------------------------------

Statistical concepts can be difficult to explain even when presented to other statisticians.  

Regardless of the target audience, we should ask ourselves questions such as: 

     1) What is the take away message for this concept? 

     2) Can the take away message be explained using an analogy or a metaphor? 

     3) Can the take away message be summarized with a catchy phrase? 

     4) What is the interpretation of the concept for the problem at hand and what are the practical implications of this interpretation for decision-making?  

When dealing with consulting clients, for instance, questions 1) and 4)  can often be the most important.  

Many statistical concepts are complex and live in an ecosystem governed by a variety of assumptions and conditions.  Case in point - the concept of p-value.  If a p-value is generated by data coming from a poorly designed/conducted study, it is essentially invalid.       

When dealing with non-statisticians, we can also ask:  What is the minimum amount of information they need to know in order for them to make sense of the information we are presenting to them and be able to use this information to make decisions? 

As an example, when we get sick, our doctor will give us a diagnosis and offer us a treatment.  We can get away with not knowing very much about the diagnosis but we definitely need to understand the treatment protocol and the results we expect from that treatment. If the treatment doesn't work as expected, we can decide to stop it and see the doctor again.   


-------------------------------------------
Isabella Ghement
Ghement Statistical Consulting Company Ltd.
-------------------------------------------

Continuing the medical analogy, it's also necessary to warn the patient/audience about
what the dangers are if the proposed prescription is not followed. It doesn't have to be
technical (e.g., something like "if you can't get a sample size of at least N, you
probably won't be able to notice a difference as small as D" instead of
"if the sample size isn't as large as N, your Type II error rate is B for delta D").

>>Kathy

-------------------------------------------
Katherine Godfrey
-------------------------------------------


Original Message:
Sent: 11-27-2014 22:26
From: Isabella Ghement
Subject: What statistical concept can you explain best to others who are not statisticians?


Statistical concepts can be difficult to explain even when presented to other statisticians.  

Regardless of the target audience, we should ask ourselves questions such as: 

     1) What is the take away message for this concept? 

     2) Can the take away message be explained using an analogy or a metaphor? 

     3) Can the take away message be summarized with a catchy phrase? 

     4) What is the interpretation of the concept for the problem at hand and what are the practical implications of this interpretation for decision-making?  

When dealing with consulting clients, for instance, questions 1) and 4)  can often be the most important.  

Many statistical concepts are complex and live in an ecosystem governed by a variety of assumptions and conditions.  Case in point - the concept of p-value.  If a p-value is generated by data coming from a poorly designed/conducted study, it is essentially invalid.       

When dealing with non-statisticians, we can also ask:  What is the minimum amount of information they need to know in order for them to make sense of the information we are presenting to them and be able to use this information to make decisions? 

As an example, when we get sick, our doctor will give us a diagnosis and offer us a treatment.  We can get away with not knowing very much about the diagnosis but we definitely need to understand the treatment protocol and the results we expect from that treatment. If the treatment doesn't work as expected, we can decide to stop it and see the doctor again.   


-------------------------------------------
Isabella Ghement
Ghement Statistical Consulting Company Ltd.
-------------------------------------------