By default, odds ratio estimates and 95% Wald confidence intervals are computed for main effects and displayed in the "Odds Ratio Estimates" table. Note that odds ratio estimates are not computed for any predictor involved in interaction or nested effects. Beginning in SAS 9.2, use the ODDSRATIO statement to compute odds ratio estimates of simple effects within an interaction as illustrated in this note. Use the CLODDS=PL option in the MODEL statement or the CL=PL option in the ODDSRATIO statement to request profile likelihood confidence intervals instead of Wald intervals. The CLODDS=BOTH option in the MODEL statement or the CL=BOTH option in the ODDSRATIO statement requests both kinds of intervals. See "Confidence Intervals for Parameters" in the Details section of the LOGISTIC documentation for details on both methods of computation.

Use the ALPHA= option in the MODEL statement to set the confidence level at something other than 95%.

You can also request custom odds ratios with the UNITS statement or the CONTRAST statement. In the CONTRAST statement, specify the linear combination of model parameters that results in a difference between two log odds and specify the ESTIMATE=BOTH option to display the estimated contrast (log odds ratio), exponentiated contrast (odds ratio), and 95% Wald confidence intervals for both. Use the ALPHA= option in the CONTRAST statement to change the confidence level. See the logistic example (Example 3) in this document for an illustration.

See "Odds Ratio Estimation" in the Details section of the LOGISTIC documentation for details on how PROC LOGISTIC computes odds ratios.

See this note and this note for why an odds ratio is not always computed as e ^β and the effects of different parameterizations in the CLASS statement.

Operating System and Release Information