05-06-2017 01:31 AM
I don't have the data set of a regression model (Model A) but only the model statistics (such as slope value, Rsquare and standard error, etc.). I need to compare the slope of Model A with the Model B that I have developed using the same dependent and independent variables that belong to different populations in both models. What would be the most appropriate test to do? Please (With sas codes if possible). Thank you.
05-06-2017 08:54 AM
I don't believe there's a default test mostly because SAS is used to process raw data nor work with summarized data.
If you have the Ns you can do the test but it would be a manual calculation.
05-06-2017 03:48 PM
Please, find attached an attachment that contains the raw data to which I had no access before. Can you please, review the SAS codes that I assembled so far to run the comparison (I don't have in-depth understanding on this topic yet). My purpose again is to compare the slopes or regression models developed by each group. Thank you again.
Can you recommend any article that covers comparison of regression models with some example or SAS codes (please, please)?
Please, feel free to ask questions.
05-06-2017 04:01 PM
05-06-2017 09:42 PM
Verify your calculation of the interaction term. I suspect it's incorrect. You can see this by looking at the N from your first regression results compared to the second regression results. SAS excludes variables with missing values.
FYI - usually I think full code reviews are beyond the scope of a public forum and won't review it, this is my personal rule. Your code is small and you can include it in posts directly, not as an attachment. Most users won't download attachments so you'll get better responses. I've also moved this to the Statistical Procedures Forum where it's more approriate.
05-06-2017 10:21 PM
For your last proc reg
proc reg data=Compare2;
model TEM = Yr_16 TEMR group_Intr;
It should be covariance analysis, try PROC GLM.
proc glm data=Compare2;
model TEM = Yr_16 TEMR Yr_16*TEMR/solution;