topic Re: help with multiple imputation in Statistical Procedures

help with multiple imputation

Kyra — Thu, 13 Sep 2018 16:23:24 GMT

Hi!

I am working with a dataset with missing values for few observations.

I would like to do multiple imputation for them.

The variables are below:

BMI_1_year Frequency Missing = 199

BMI_6_mon Frequency Missing = 122

Post_op_BMI_4_weeks Frequency Missing = 127

Pre_op_BMI Frequency Missing = 177

Please let me know sas code to do it.

I have attached a part of my dataset.

Thanks,

Re: help with multiple imputation

Kyra — Thu, 13 Sep 2018 16:25:18 GMT

I am not able to attach my dataset. sorry. It says "The contents of the attachment doesn't match its file type."

Re: help with multiple imputation

SAS_Rob — Thu, 13 Sep 2018 16:26:43 GMT

You will need to decide which methodology you want to use specifically. This link in the MI documentation will help you to decide:

https://go.documentation.sas.com/?docsetId=statug&docsetVersion=14.3&docsetTarget=statug_mi_details05.htm&locale=en

Once you decide between MCMC vs FS/Monotone then there are good examples of the coding in the documentation as well.

https://go.documentation.sas.com/?docsetId=statug&docsetVersion=14.3&docsetTarget=statug_mi_examples.htm&locale=en

Re: help with multiple imputation

Kyra — Thu, 13 Sep 2018 16:33:31 GMT

Thank you very much for the reply. I will go through the paper.

Re: help with multiple imputation

Kyra — Thu, 13 Sep 2018 16:52:39 GMT

How do i know if the missing pattern is monotone or arbitrary?

I looked into the examples provided in document but it is not clear.

Thanks.

Re: help with multiple imputation

SAS_Rob — Thu, 13 Sep 2018 17:03:31 GMT

MI provides a table in the output with the missing data pattern. Since you have all continuous variables you could use the MCMC method and the NIMPUTE=0 option to see only the missing data pattern.

Re: help with multiple imputation

Kyra — Thu, 13 Sep 2018 17:30:46 GMT

I used the following code:

proc mi data=red.sleeve1 nimpute=0 ;
mcmc chain=multiple displayinit initial=em(itprint);
var BMI_1_year BMI_6_mon Post_op_BMI_4_weeks Pre_op_BMI ;
run;

The sas output i got is attached. How do i interpret this. Thanks.