Question: the multiple linear regression analysis using PROC SURVEYREG and testing for interactions

[berylsky1]

Hi all,

I am doing multiple linear regression using proc survreg and want to test interaction between stressors and gender. I want to know if my codes below are correct for the model 1 and model 2. I really appreciate any feedback.

This was a list of variables I used for the analysis.

Gen2_cat= gender (binary)

ethnicity_group= ethnicity (4 categorical)

maritalstatus_dummy=marital status (binary)

nativity2=nativity (binary)

COVID_ImpactOverall_new= continuous

FAsocitalracism= continuous

Socialstressors4= continuous

materialstressors4 = continuous

Age_BirthYear_Calculate= continuous

miss_binary('1') = complete sample

/*Model1: Interaction Stressors*Gender – Interaction Testing*/
proc surveyreg data=mental8 alpha=0.05; /*alpha-set alpha level*/
 class gen2_cat (ref="0") ethnicity_group (ref="0") maritalstatus_dummy (ref="0") nativity2_cat (ref="0"); /*reference*/
 model COVID_ImpactOverall_new= FAsocitalracism socialstressor4_v1 materialstressors4  FAsocitalracism*gen2_cat socialstressor4_v1*gen2_cat materialstressors4*gen2_cat Age_BirthYear_Calculate gen2_cat ethnicity_group maritalstatus_dummy nativity2_cat income_new_n edu5_cat_n/solution CLPARM STB ADJRSQ;/*CLPARM=display 95%CI*/
 weight WEIGHTS_A7;
 domain miss_binary('1');
run;

The interaction between stressors and gender I got :

FAsocitalracism*gen2_cat : p=0.3537

socialstressor4_v1*gen2_cat : p=0.015

materialstressors4*gen2_cat : p=0.486

Since interaction social stressor and gender was significant, I would like to stratify the model by gender. In this case, should I include all stressors or just social stressor in the stratified model?

/*Model2: Stratified by Gender*/
proc surveyreg data=mental8 alpha=0.05; /*alpha-set alpha level*/
 class gen2_cat (ref="0") ethnicity_group (ref="0") maritalstatus_dummy (ref="0") nativity2_cat (ref="0"); /*reference*/
 model COVID_ImpactOverall_new= FAsocitalracism socialstressor4_v1 materialstressors4  Age_BirthYear_Calculate ethnicity_group maritalstatus_dummy 
 nativity2_cat income_new_n edu5_cat_n/solution CLPARM STB ADJRSQ;/*CLPARM=display 95%CI*/
 weight WEIGHTS_A7;
 domain miss_binary('1')*gen2_cat; /*stratify by gender*/
run;

Also, I want to know how I can get predicted mean mental health by stressors (i.e., FAsocitalracism, socialstressor4, materialstressors4) for female and male in proc surveyreg and create a graph based on the results (like the graph I attached below). 

I would appreciate if you could help me with this! Thank you in advance!

Looking forward to your response!!!