Handling missing values in PROC SURVEYLOGISTIC for multiple regression model

tolesa2003 — Sat, 25 May 2019 10:13:35 GMT

Currently I'm analyzing complex survey design data. I have missing values for most of my variables due to recoding form non-responses such as 99, 999, 9999 etc. When I fit the multiple model using CLUSTER, STRATUM and PERWEIGHT options the model was severely affected due to missing values. How can I handle the missing and fit the model with the following?

PROC SURVEYLOGISTIC DATA= <hjGG>

WEIGHT PERWEIGHT;
Cluster PSU;
Stratum Domain;

Class <abcm>;

Model hhchc=<gahdj>;

RUN;

Re: Handling missing values in PROC SURVEYLOGISTIC for multiple regression model

ballardw — Tue, 28 May 2019 17:51:46 GMT

One option is to use MISSING on the Proc statement. Then "missing" becomes a category.

You don't say if the missing are occurring with your class variables, other variables or both or even how many variables are in your model. It might be that you have too many variables for your data.

topic Handling missing values in PROC SURVEYLOGISTIC for multiple regression model in Statistical Procedures

Handling missing values in PROC SURVEYLOGISTIC for multiple regression model

Re: Handling missing values in PROC SURVEYLOGISTIC for multiple regression model