Pick the Tools that Can Answer the Question Posted on November 15, 2007 Eric A. Youngstrom (bio) shares his enthusiasm for clinically-relevant measurements.

Talking about career paths and also talking about measurement and statistics, I don’t think that I ever would have guessed that I would be as enthusiastic about measurement and assessment as I am now. The first stats course that I took as an undergraduate was my lowest grade as an undergraduate. It bored me to tears. It was painful to get through.

The second course that I took was a little bit better, and then the third course which I took, something clicked and I suddenly realized that what I was learning was a set of tools or a set of skills that would let me answer the questions that I really cared about. And ever since then, the way which I’ve approached this has been, “Think about the interesting question, focus on the question and then pick the tools that will let you answer the question the way that you want.”

So concretely, everyone learns how to do a t-test. T-test, you sort people into two groups: these are the people with bipolar disorder; these are the people without. And then you go ahead and you compare them on something.

Well, that test is exactly backwards from the way that clinicians need to operate. You would bring your child into me and I get to look at one child. I can pick whatever test I want, but no one’s going to tell me, “Does this child really have bipolar disorder or not?” There’s no bat phone or hotline to the higher power. No one’s ever taken my caseload for me and told me the truth.

So clinicians have to operate exactly the opposite direction. We can pick the tests, we look at it, it comes back, child scores an 87. Okay, now what? It turns out that there are a whole family of statistics that I was never taught in school. They’re not part of the standard psychological or psychiatric training, but what they focus on is exactly that question approached that way: “Here’s a test result, what’s it trying to tell you about what’s going on with the person?”

And so that understanding, that light bulb coming on, is what’s made me passionate about learning to do logistic regression, to do ROC analysis, to do things that you see reflected in the papers coming out of our group now.

And for me, I get excited about it talking with research audiences, but I get excited talking about it to clinical audiences, too, because there’s this perceived gap between science and practice, and part of it is that the practitioners look at these articles and see lots of t-tests or ANOVAS and they’re like, “That doesn’t tell me what to do with the next patient.”

But there are statistics that do help us to understand, “Here’s what I should be doing with the next patient.” And so I’ve really caught fire about trying to learn those tools, share those tools with the people collaborating with me, but also with anyone else who will hold still.

« Back to Article