Compare two non-nested models

[UcheOkoro]

Hello, 

Please, I need to compare two non nested  models. Please, I need help with SAS code for Monte Carlo simulation in structural equation modeling to do this comparison.

Thank you

[Rick_SAS]

As a starter, you need to know how to simulate multivariate normal data and data for a linear regression model. If you have a moderately complicated model with multivariate relationships, I suggest you use SAS/IML software rather than the DATA step in Base SAS. Here are some tips to get you started:

• How to generate MVN data with SAS/IML   (If you don't have SAS/IML, you can use PROC SIMNORMAL.) 
• How to generate data for a LOGISTIC regression model in SAS/IML (just omit the logistic transformation to get a linear model). If you don't have SAS/IML, you can use the DATA step to simulate regression data.)
• General tips for data simulation in SAS

SEM models are fairly complicated, so be sure to master the basics before you attempt SEM.

[UcheOkoro]

Thank you so much for your response. I would like to compare the following two non-nested models. I do not want simulate a data set. Please, what is the best way to do this?

/****MODEL WITH SEPSIS, SURGERY AND ED_DISPO*****/
   PROC LOGISTIC DATA=Averasepsis_data10C3b OUTEST=_LNLIKE4_ DESC;
  CLASS  triage_pulse2(ref='<96') Lactate1_cat(ref='<2')  mode_arrival(ref='private tranp') bmi2(ref='<18.5') total_sofa1_ed2(ref='<4') pmh_copd(ref='No') 
pmh_cirrhosis(ref='No')  pmh_transplant(ref='No') pmh_cancer(ref='No') Age2(ref='<30') pmh_diabetes(ref='No') pmh_dialysis(ref='No') infection_ed_source3(ref='resp_pnuem')
 sex2(ref='Male') RUCA2(ref='Two') ed_provider_type2(ref='Physician') ICU_yn(ref='No') sepsis_septic_diag(ref='No') sepsis_septic_diag(ref='No')ed_dispo(ref='Admit') surgery(ref='No');
	MODEL telemed_use (Event='Yes') =  triage_pulse2 Lactate1_cat triage_systolic mode_arrival delta_SOFA bmi2 citypop_2019
EDVolume_2019 totalbeds transferdist sex2 RUCA2 ed_provider_type2 ICU_yn sepsis_septic_diag ed_dispo surgery/lackfit rsquare;
  RUN;

  /****MODEL WITH SEPSIS and SURGERY*****/
   PROC LOGISTIC DATA=Averasepsis_data10C3b OUTEST=_LNLIKE5_ DESC;
  CLASS  triage_pulse2(ref='<96') Lactate1_cat(ref='<2')  mode_arrival(ref='private tranp') bmi2(ref='<18.5') total_sofa1_ed2(ref='<4') pmh_copd(ref='No') 
pmh_cirrhosis(ref='No')  pmh_transplant(ref='No') pmh_cancer(ref='No') Age2(ref='<30') pmh_diabetes(ref='No') pmh_dialysis(ref='No') infection_ed_source3(ref='resp_pnuem')
 sex2(ref='Male') RUCA2(ref='Two') ed_provider_type2(ref='Physician') ICU_yn(ref='No') sepsis_septic_diag(ref='No') sepsis_septic_diag(ref='No') surgery(ref='No');
	MODEL telemed_use (Event='Yes') =  triage_pulse2 Lactate1_cat triage_systolic mode_arrival delta_SOFA bmi2 citypop_2019
EDVolume_2019 totalbeds transferdist sex2 RUCA2 ed_provider_type2 ICU_yn sepsis_septic_diag surgery/lackfit rsquare;
  RUN;

[Rick_SAS · Accepted solution]

I'm confused. The title of your post is "Monte Carlo simulation in structural equation modeling in SAS." But now you are asking about how to compare two logistic models.  

If you want to compare two logistic models, one way is to compare the area under their ROC curves (C statistic). See https://communities.sas.com/t5/Statistical-Procedures/how-to-compare-two-models/td-p/121251

For other options, see https://www.lexjansen.com/mwsug/2016/AA/MWSUG-2016-AA19.pdf

[UcheOkoro]

Thank you so much, Rick_SAS. This was very helpful!