Statistical Consulting: Telling People What They Don’t Want to Know

Being a Type-AAA personality,  in addition to running the Julia Group, I have a ‘day job’ as a statistical consultant at a university where the communications people shudder as they walk by me. (I love the title of the book Molly Ivins Can’t Say That Can She? Simply because it reminds me of the reactions I get just going through life.) Hence, this blog is hosted by The Julia Group as Julia could care less what I say as long as I pay her phone bill enabling an unending stream of text messages to everyone short of God, whose cell phone number she has been unable to learn – yet.

Oh, speaking of The Julia Group, we’re booked through September, 2010. I am doing as much work as I want to do and I am not interested in hiring you and getting more work for you to do. Don’t call me or ask me to be your friend on linked in or any of those things. If you do need statistical consulting after September, 2010, though, look me up.

Where we were – oh yes, statistical consulting at the university. Much of job entails telling, or at least hinting, things to people that they don’t want to know. Before that I was a professor teaching statistics to people who didn’t want to learn it.

A nice part of my job, though, is I get paid to read books and look things up on the Internet. As my grandmother told me about my post-doctoral research, which was essentially the same thing, with some journal article writing thrown in, and at much lower pay,

“Mija, reading books and talking about them is not a job. It is what people do after their job to relax. You better go back to whoever hired you at that university and ask. There’s got to be more to it than that.”

Here is one thing people don’t want to know – when you glance at your data, find two groups have a substantially different mean and then do a planned comparison as if those are the only two groups you ever thought of comparing, what you have REALLY done statistically speaking is all possible pairwise comparisons and then selected out the two groups that were different. You really ought to use the Tukey or other test that does all possible comparisons, suck it up and take the error rate and level of significance that goes with what you really did. The fact that you ‘eyeballed’ it versus running the Tukey first doesn’t change anything.

But you’re not going to listen to me, are you? No-o-o , you are going to go ahead and do a t-test between those two groups aren’t you? Fine! Don’t say I didn’t warn you.

On a positive note, though, when I run across a website that talks about statistics, it is like finding a friend. Yesterday, I was looking for a book on statistics with SPSS and I found 42 “Introduction to — ” and “Getting started with — ” all the way back for 12 versions. I cannot imagine what I need less than a book entitled “Getting started with SPSS 8.0”.  A mushroom brush, maybe?

So, I was very happy to find this website by David Howell where he discussed multiple comparisons with repeated measures, among other things.  It wasn’t what I was looking for, really, but I think most of what I have learned in life occurred when I was actually looking for something else.