We use cookies to help give you the best possible experience on our site. Strictly necessary and functional cookies support login and shopping cart features, they cannot be disabled. Performance cookies support site performance analysis. These are optional and will be disabled if you click on Reject.
By clicking Accept you agree to our use of Performance cookies as detailed in our Privacy Policy.
Accept Reject
1. What is GENMOD? PROC GENMOD fits generalized linear models (GLMs) . It extends ordinary linear regression to response variables that have error distributions other than normal (e.g., binomial, Poisson, gamma). It also handles repeated measures via GEE (Generalized Estimating Equations).
PROC GENMOD DATA=mydata; MODEL y = x / DIST=BINOMIAL; OUTPUT OUT=preds PRED=pred_prob; RUN; | Mistake | Consequence | Fix | |---------|-------------|-----| | Forgetting CLASS for categorical vars | Treated as continuous | List all categorical vars in CLASS | | Using TYPE=AR(1) with unbalanced times | Invalid correlation | Use TYPE=UN or EXCH with CORRW | | Ignoring overdispersion | Inflated significance | Try DIST=NEGBIN or SCALE=DEVIANCE | | Not exponentiating coefficients for ratios | Misinterpretation | Use EXP option in ESTIMATE or LSMEANS | 8. Comparison: GENMOD vs. Other SAS Procedures for GLMs | Feature | GENMOD | GLIMMIX | LOGISTIC | REG | |---------|--------|---------|----------|-----| | Normal errors | Yes | Yes | No | Yes | | Binomial | Yes | Yes | Yes | No | | Poisson/NegBin | Yes | Yes | No | No | | Random effects | No (GEE only) | Yes | No | No | | GEE for repeated measures | Yes | No | No | No | | LSMEANS on link scale | Yes | Yes | Limited | No | genmod
