topic Proc glm pairwise deletion in Statistical Procedures

Proc glm pairwise deletion

owiseilayyan — Sun, 11 Jun 2017 20:36:10 GMT

Hello,

I want to run a regression model using proc glm. I have missing valuse on both dependent and independent variables. The defult in SAS is listwise deletion but I want to do pairwise deletion. Howver, I couldnt find a statement to do pairwise deletion.

Could anyone help me in that?

Thank you,

Owis

Re: Proc glm pairwise deletion

Rick_SAS — Mon, 12 Jun 2017 00:56:04 GMT

GLM does not delete any observations for which the Dependent Variables (DBV) is missing. It deletes any observation for which one or more Indep Variables (IVs) are missing. Thus "pairwise deletion" doesn't make sense in the general setting.

What models are you trying to fit? A single model that has one IV? A series of models? Give us sample PROC GLM statements.

Re: Proc glm pairwise deletion

owiseilayyan — Mon, 12 Jun 2017 04:16:35 GMT

Thank you for your reply,

Here is my proc glm:

proc glm data= CRCT_DATAI;
class gender (ref = 'Woman') cGPA (ref = '3.3-3.7') Lev_educ (ref = "Bachelor's degree") Geo_setting (ref = 'Urban') Focuscare (ref = "General musculoskeletal care (spine and extremities)") style (ref = 'Pragmat');
model KABQ_Knowledge = age gender cGPA Lev_educ Geo_setting Focuscare style/ solution ss3 ;
run;

The output says:

Number of observation read: 127

Number of Observation used: 49

However, when I used simple regression, almost all data was used!

Thank you,

Owis

Re: Proc glm pairwise deletion

Reeza — Mon, 12 Jun 2017 05:49:03 GMT

What's a simple regression?

Why do you have so many missing, it's almost 2/3 of your data?