topic Re: How to use available case analysis in SAS Procedures

How to use available case analysis

npandis — Thu, 21 Jun 2018 13:28:13 GMT

Hi,

I have a dataset with approximately 1300 observations and I am interested in fitting a series of models using:

proc logistic

proc surveylogistic

proc genmod

Some of the included covariates have missing values and the missing values of the covariates belong to different observations. What I am trying to say is that for a single observations not all covariates included in the model are missing.

When I fit the models I get about 15% of the observations not used.

Is there an option to tell SAS to include the observations in the analysis which have at least one covariate with no missing data for the observation (to use available case analysis)?

Thank you,

Nikos

Re: How to use available case analysis

ballardw — Thu, 21 Jun 2018 14:24:18 GMT

If the variables you are concerned with are CLASS variables you can use the option MISSING on the class statement to indicate that missing is considered a valid level in Logistic and Genmod. I don't see an equivalent for Surveylogistic though.

I don't believe there is another option if the predictor is not a class variable that might help directly.

Another option might be to impute values for the missing variables.

Re: How to use available case analysis

npandis — Thu, 21 Jun 2018 15:53:03 GMT

Thank you and best wishes,

nikos