https://communities.sas.com/t5/Statistical-Procedures/Advice-for-calculating-predictive-margins-with-Zero-Inflated/m-p/753812#M36679 <P>If the issue is that you have a large proportion of zero values, exceeding what the Poisson distribution allows, then you can try the negative binomial or the Tweedie distribution. The Tweedie distribution can handle large point masses at zero.&nbsp;</P> Tue, 13 Jul 2021 16:25:06 GMT StatDave 2021-07-13T16:25:06Z https://communities.sas.com/t5/Statistical-Procedures/Advice-for-calculating-predictive-margins-with-Zero-Inflated/m-p/753762#M36677 <P>I can see in the documentation on&nbsp;<A href="https://support.sas.com/kb/63/038.html" target="_blank">https://support.sas.com/kb/63/038.html</A>&nbsp;that the Margins Macro does not work for zero-inflated models. Grateful for any advice on alternative approaches for ZIP models!</P><P>&nbsp;</P><P>Excerpt from 63038:</P><P><EM><STRONG>LIMITATIONS:&nbsp;</STRONG>The Margins macro can be used only with a subset of the models available in PROC GENMOD since that procedure is used to fit the specified model. It cannot be used with multinomial or zero-inflated models available in GENMOD...</EM></P> Tue, 13 Jul 2021 13:00:06 GMT https://communities.sas.com/t5/Statistical-Procedures/Advice-for-calculating-predictive-margins-with-Zero-Inflated/m-p/753762#M36677 tymon 2021-07-13T13:00:06Z https://communities.sas.com/t5/Statistical-Procedures/Advice-for-calculating-predictive-margins-with-Zero-Inflated/m-p/753812#M36679 <P>If the issue is that you have a large proportion of zero values, exceeding what the Poisson distribution allows, then you can try the negative binomial or the Tweedie distribution. The Tweedie distribution can handle large point masses at zero.&nbsp;</P> Tue, 13 Jul 2021 16:25:06 GMT https://communities.sas.com/t5/Statistical-Procedures/Advice-for-calculating-predictive-margins-with-Zero-Inflated/m-p/753812#M36679 StatDave 2021-07-13T16:25:06Z