ASA Connect

Hi Eric,

The answer to your question depends on many variables.

I'd like to think that we all start as consultants and, as we mature professionally, we become valued and trusted collaborators.  

Throughout our careers, we most likely oscillate between the consulting and collaboration poles, depending on the clients we work with and the scope/complexity of the problems we need to solve. 

The real question is: How can we add more value to our clients, regardless of the context in which we operate? 

In my experience, the value that we can add is often non-statistical in nature and is generated by our soft skills.  

Looking forward to seeing what other colleagues have to say.

Isabella

Eric and I have discussed this before, and as he knows, I disagree strongly with his definitions. I think they are wrong and harmful to our profession. Alas, I am booked this morning so don't have time to respond fully. Here's a short version.

Statistical consulting needs to answer the clients' questions too; otherwise it's just bad statistical consulting. One might even say it's not really statistics at all when the data is removed from its context. To define statistical consulting in this way normalizes this poor practice.

Additionally, to use these terms as definitions, as Eric does, means that any of us who have said we do statistical consulting are included in this category, which is false. Eric, I feel insulted every time you use the words this way. I would prefer that you say that "we should call this statistical consulting, and that statistical collaboration."

Finally, I don't believe the words are actually used this way in practice. I interviewed several of my collaborators recently about this, and none of them expressed anything close to this. The most common distinction was about the level of the involvement; a consultant could be someone you just see once or twice to get advice and guidance, but a collaborator is someone who is fully involved and has an equal stake in the result.

Now, I do believe that the word collaboration is viewed more favorably, and so have started to use it more myself, even for small projects, but I do not believe we should elevate this word by degrading the phrase statistical consulting.

I agree with Aaron. People often come to see a statistical consultant  saying that they have a statistics problem, but often the solution to that perceived statistics  problem doesn't really solve their substantive research problem. Every statistical consultant has to make sure that (s)he addresses the research or business problem facing the client. Sometimes that generates an entirely different way of thinking about the research problem. Competent consulting demands just as much context as competent collaboration.

If I had to draw a distinction, it would be on the depth of the relationship - just the way Aaron describes it.

The distinction lies in the definitions of the words "consult" and "collaborate." From Merriam-Webster:

collaborate - to work jointly with others or together especially in an intellectual endeavor

consult -  to ask the advice or opinion of

So I think Aaron's collaborators have the right idea that the difference between statistical consulting and statistical collaboration is the level of involvement and not the type of question being answered.

The difference between statistical consulting and statistical collaboration seems similar to the difference between authorship and an acknowledgement.  Many universities have a statistical consulting group.  I am wondering if the statistical community would be more successful in outreach is we offered statistical collaboration services rather than consulting services?

I agree with Dena Schanzer with the exception that "more" collaboration should be offered as opposed to "rather than consulting services".  Guidelines or awareness as to how to recognize when "consulting" starts to turn into "collaboration" might be helpful as this can also lead into ethical concerns if the expectations differ between the statistician and client/scientist/manager.  And one needs to recognize that the fastest way to lose a client is to say "I'll help you but only if I become a co-author."

Thanks, Eric, for bringing this up for discussion.  I think this is a conversation that many of my colleagues (bio/statisticians in schools of medicine) have on a weekly basis.  

First, I want to comment on the distinction between consulting and collaboration.  I think your definitions are spot on, especially as they relate to the amount of time, energy, and investment a statistician may put into a particular project.  And I think many (if not most) in our profession are clear on the differences between consulting and collaboration.  

However, the problem lies in the perception of consulting/collaboration by our non-statistical colleagues/investigators.  I don't want to paint with a broad brush (since I have many collaborators who are fantastic), but I think many investigators view what we do solely as consulting when in fact it should be collaboration.  This includes, but is not limited to, last-minute power calculations on grant proposals, not including the statistician on any project-related conversations until it's too late (i.e. year 5 of the grant), etc.  

Furthermore, there are many investigators who truly want to collaborate with statisticians, but their expectations of what "collaboration" means falls short.  For example, their assumption is that for a little bit of effort (i.e 5%), they will have a PhD level statistician who will do all of the data management, statistical programming & analysis, and writing of papers.  While this is true to some degree, none of this can be done well alone and without high-quality statistical and data management staff (bachelors/masters level who also come at a well-justified cost).  This is even more vital in clinical trials.  

As a statistical community, we have to continually make sure we're advancing our profession forward.  Not only do we have to make the case to our colleagues/investigators that collaboration (versus consultation) is vital to the success of their research projects, but we also have to make the case that it's vital to the success of our careers.  Namely those of us in academia.

Thanks again, Eric, for stimulating discussion on this.  

--kaleab

AS a long term consultant (40+ years), I have always thought that any consulting gig should also include collaboration. It sounds like most of the respondents here are in Academia, which may have significantly different experiences than me.

I find it is extremely important to understand the business problem of my clients and it is important to give them a feel for what I look for from the statistical or analytical side. They need to know why sample size and selection is important and I need to learn from them potential biases in the data and any potential drivers for the results. As the project matures I have always asked the client to review results, often having to explain the implications of those results.

I have to admit that getting "collaboration" from the client is sometimes like pulling teeth as they often just want a result.

OTOH, I often think of collaboration as more of a brainstorming exercise where I will work with other statisticians or analysts to review work to get a different POV, so I like to think of "consulting" as exercising a specific expertise, but that does not exclude the kind of collaboration discussed above.

Hello everyone,

I've been reading theis thread with interest and want to add in my take on the issue. First, I think we can all agree that a consultant collaborates with his or her client and a collaborating statistician provides consultation to the client. But the degree of interaction between statistician and client may be the key to differentiating the two terms. 

To me, a consultant is generally someone who is technically capable and can help a client work through an existing project with little or no input into the formulation of the project. When the work is completed, the consultant moves on to a new client. A collaborator is someone who has developed a relationship, likely long-term, where teamwork, trust and communication are critical components and who works together with the client to develop the project to everyone's satisfaction before doing all the technical stuff. An analogy might be the difference between fighting a fire and taking action to prevent a fire from starting. 

I work in industry providing statistical services to my clients. My title and company name suggests I am a Consultant, and I certainly offer that service, but I always strive to "Collaborate" with every organization I deal with. Giving the client the extra input and effort can lead to a one-time bit of work become something more.

From my experience, a consultant is the person you pay to talk with after you messed up your experiment. The collaborator is paid by the grant before you mess up your experiment;-) The consultant walks away with money. The collaborator gets a publication. 

I have heard of industrial statistical consultants, academic statistical consultants and academic collaborators.

If you look at a lot of the universities in Michigan, they tend to have a group dedicated to statistical consulting and another group dedicated to collaboration. The consultants seem to be there for a short term gain, be graduate students or have MS degrees. The collaborators tend to be PhDs and seem to be there for a longer duration.

In industry, everyone seems to be a consultant. I've also found that industrial consultants tend to be skilled in an area of expertise that's not statistics and have some fundamental knowledge of statistics, especially Design of Experiments. 

In the Metro Detroit area, we have CSCAR at The University of Michigan and RDA at Wayne State. Both groups consider themselves statistical consultants. CSCAR has MS and PhD level statisticians working there. As someone that has tried to use their services, I was shocked by how little they know about the science they are helping with. They (falsely) assume that the scientists already has a good grasp on designing experiments. (Most scientists know that you can't change more than one thing at a time during an experiment and feel multiple t-tests are the best method for analyzing more than 2 groups... SERIOUSLY!!!!)

As far as I can tell, RDA is staffed by psychology graduate students who might have taken 2-3 stats classes from the psychology department. I have seen some of the experiments and analyses these folks have done. (They would have failed traditional stats classes.) I even got into an argument with one over whether or not Factorial Designs exist!!! (Since it doesn't appear in their text books, it must not exist. They also knew that Plackett Burman Designs didn't fit with what my friend wanted to do, even though they never heard of them. They also felt the Design of Experiments textbook was wrong, even though the author designed the same type of experiment.)They also charged my friend quite a lot for their garbage.  

I also know that both Wayne State and U of Mich have lots of statistical collaborators in their biostatistics departments. Through ASA and my former work, I have met several biostatistical collaborators at U of Mich and am again shocked at how little they know about the science they are working on. I have also found that they are afraid to find things out. They also tend to take the word of the scientist/researcher as the gospel, and tend not to question the scientists assumptions. (i.e. They don't use "doctor" as a variable in their studies. All doctors are 'equally' good and so are their measuring devices... cuz they said so.)

Having talked to some of the collaborators at Wayne State, like Robert, I get the feeling they are not so naive. I also get the feeling they don't/won't have enough collaborators once the new integrative biosciences center is up and running at full capacity. 

In my very limited experience, the distinction seems to me to be partly contractual.  I have done unpaid collaborations and others as a part of my job.  When I have a contract specifically for a project, that is consulting.  When I give advice, that is consulting. The last entails no collaboration.  Perhaps we need a finer terminology, as I used consulting in two different senses.  Let me give that a shot.

                      Collaboration              |    No collaboration

---------------------------------------------------------------

Contract      |   Consulting                 |  Advising

No contract  |   "Pure" collaboration  |  Advising

Note:  "No contract" includes as part of a regular work arrangement

According to this classification, when I "consult with" someone about some work, I'm really advising.

Frankly, I think this whole thing is a bit of picking at nits.

What is important is that when working with others, it is kind of like the book "All I Really Need to Know I Learned in Kindergarten" by Robert Fulghum.

I have worked in the past a consulting statistician at The University of the West Indies.

My experience is that there an entire spectrum between the polar opposites of consultant and collaborator. I recall that I found the project to be much more satisfying if I were a collaborator and got joint authorship of a paper at the end of the exercise. Of course, consulting fees were always welcome.

To state what is probably obvious, statisticians have to be flexible. They need to to have a good idea of what level of collaboration would be involved in a new project. Some clients do have very good knowledge ( based on previous projects ) of what they need and why a factor analysis would be better than a multiple regression in a given situation. Less experienced clients may be shaky in their knowledge of what t-tests or ANOVA do. Still others may not realize that data cleaning and data management is a big part of the project.

To me, a major distinction is the influence you have as a collaborator but not as a consultant. As a consultant, I perform advisory services. At the end of the day the client can completely disregard my recommendations and do what they want to do, as ill-advised as that may be. As a collaborator, I have direct influence on the outcome.

I think this somewhat ties into the ideas I have been seeing here. Contributing to a paper as an author gives me direct access to the outcome. Being involved early in the process allows me to shape and contribute towards the end product. Coming aboard late in the game just to explain some statistics to the client for them to digest and somehow incorporate into their own work does not.

Sharing with you the latest AmStat News article published this month:

A Recipe for Successful Collaborations: A JSM 2015 Panel Discussion

http://magazine.amstat.org/blog/2016/01/01/a-recipe-for-successful-collaborations/