Confusion regarding multivariate and controlling

[sasnewbie12]

Hi

I am conducting a study comparing two treatments. What difference will it make if I use age in the multivariate analysis to control for any confounding effects age (<=18 or >18) will have versus doing two different studies (one in adult patients only and one in pediatrics patients only). 

Thank you

[Reeza · Accepted solution]

Entirely separate analysis means that the variance is calculated differently whereas they're included in the model. Instead of separating it out entirely, you can also consider a STRATA variable. 

So your options are :

1. Separate models (BY statement)

2. STRATA statement - variance for both are considered

3. Model entry - variance considered, interaction terms possible. 

[PaigeMiller]

As I see it, age should be treated as a continuous variable, because it is a continuous variable. If you do two different studies, you are now analyzing the data as if age was a discrete variable. Furthermore, someone who is 17 years 11 months is treated by the analysis very differently than someone who is 18 years one month. So, with all that in mind, I think the model would have greater predictive power if you treat age as a continuous variable. And not only that, the sample size used in the modeling is greater if you treat age as a continuous variable in one large analysis.