Unit D Translation Activity

These activities are for practice only — there is nothing to turn in.

Units B and C asked you to translate a study into a model. This one asks you to translate a problem into a diagnosis and a response. Each scenario below describes something that went wrong, or might have, in the words a collaborator would use.

For each one, answer:

1. Which assumption or data problem is being described?
2. Which diagnostic would you run to confirm it, and what would confirming it look like?
3. What would you do about it — and what would you have to report either way?

Example 1

My residual plot has a clear funnel shape: predictions are tight at the low end and fan out badly at the high end. My supervisor says the p-values are fine because the sample is large.

Example 2

I measured self-esteem with a three-item scale that turned out to have an alpha of .55. My predictor barely reached significance and the effect looks much smaller than the published literature suggests.

Example 3

About 18% of my participants skipped the income question. I dropped those rows so the model would run. My reviewer wants to know whether that was defensible, and I am not sure how to answer.

Example 4

My outcome is badly skewed and I have 40 participants. A colleague says to log it, another says to run a nonparametric test, and a third says to bootstrap the confidence intervals. They cannot all be right.

Example 5

Two observations have Cook’s distance far above everything else. Removing them flips my main effect from significant to not. I would rather not remove them, but I also do not want to report a result that hangs on two people.

Bring your own

Bring a diagnostic plot from your own data — ideally one you were not sure how to read — and work through the same three questions with someone else.