Logan Hanks
Essays and newsletter on research, science communication, and grad-school life.
Reviewers accept manuscripts they can trust, and trust comes from a clearly specified causal question, a defensible design, and honest sensitivity analyses—not from a bigger dataset or a flashier model.
The odds ratio is the workhorse of logistic regression, but it routinely overstates how much an exposure matters when the outcome is common. Here is how to read it honestly.