Statistics collaboration As a biologist, many of the research projects I work on at some point involve collaboration with a statistician, whether it be for simple advice or for implementing and testing a model for my data. My statistics colleagues admit that they do a significant amount of collaboration, insomuch that the tenure review process only considers papers on which they are the first or last author.
What would make me (or any other scientist) a better collaborator? What would make it easier for you (as a statistician) to work with me? Specifically, what is one statistics concept you wish all of your scientist collaborators already understood?
 A: To get a good answer, you must write a good question. Answering a statistics question without context is like boxing blindfolded. You might knock your opponent out, or you might break your hand on the ring post.
What goes into a good question?


*

*Tell us the PROBLEM you are trying to solve. That is, the substantive problem, not the statistical aspects.

*Tell us what math and statistics you know. If you’ve had one course in Introductory Stat, then it won’t make sense for us to give you an answer full of mixed model theory and matrix algebra. On the other hand, if you’ve got several courses or lots of experience, then we can assume you know some basics.

*Tell us what data you have, where it came from, what is missing, how many variables, what are the Dependent Variables (DVs) and Independent Variables (IVs) – if any, and anything else we need to know about the data. Also tell us which (if any) statistical software you use.

*Are you thinking of hiring a consultant, or do you just want pointers in some direction?

*THEN, and ONLY THEN tell us what you’ve tried, why you aren’t happy, and so on.
A: Having no preconceived ideas about the method you should use solely based on papers. Their ideas, logic or methods may be faulty. You want to think about your problem and use the most appropriate set of tools. This reminds me of reproducing cited information without checking the source.
On the other hand, paper with methods (or logic) that differs from the rest of literature may hinder or cull a review process because "it's not the norm".
A: My answer is from the point of view of an UK academic statistician. In particular, as an academic that gets judged on advances in statistical methodology.

What would make me (or any other
  scientist) a better collaborator? 

To be blunt - money. My time isn't free and I (as an academic) don't get employed to carry out standard statistical analysis. Even being first/last author on a paper that uses standard methodology is worth very little to me (in terms promotion and my personal research). Paying for my time will buy me out of administrative or teaching duties. Payment could be through a joint grant.
In the UK, every five or so years academics have to submit their four best papers. My papers are judged on their contribution to the statistical literature. It sucks, but that's the way it is. 
Now it may well be that you have a very interesting problem which would lead to advances in statistical techniques. However, just think about the size of your statistics department compared to the rest of the Uni. There probably won't be enough statisticians to go around.
In saying that, I do try and do some "statistical consultancy" once a year to broaden my interests and to help for teaching purposes. This year I did some survival analysis. However, I've never advertised this fact and I still get half dozen requests each year for help!
Sorry for being so negative :(

Specifically, what is one statistics
  concept you wish all of your scientist
  collaborators already understood?

That statisticians do statistical research. As one of my collaborators said:

Surely there's nothing left to solve in statistics?

A: I think the concept  that few scientists grasp is this: A statistical result can really only be taken at face value when the statistical methods were chosen in advance while the experiment was being planned (or while preliminary data were collected to polish methods). 
You are likely to be mislead if you first analyze the data this way, then that way, then try something else, then analyze only a subset of data, then analyze only that subset after removing an obvious outlier..... and only stop when the results match your preconceptions or has lot of asterisks. That is a fine way to generate an hypothesis, but not an appropriate way to test one.
