Binary and count outcomes require regression models that match the outcome type. Logistic regression is used for binary outcomes, such as response, disease status, or event occurrence, whereas Poisson or negative binomial regression is used for count outcomes, such as numbers of adverse events, hospital visits, or infections.
| Outcome | Model | Main effect measure |
|---|---|---|
| Binary | Logistic regression | Odds ratio |
| Count | Poisson regression | Count ratio |
| Overdispersed count | Negative binomial regression | Count ratio |