Get different results for one parameter t value when using proc glm and proc reg

[Tobias80]

Hello,

I have a continuous explanatory variable, a continuous dependent variable, a dummy variable and an interaction variable. However the t value for the variable mentioned lnx gets different results for the two methods. Does someone have an idea of why?

The overall F-test is the same for both proc reg and glm.

The dummy (dummy_APO_E ) variable is either 0 or 1 and the interaction (interaktion) variable is the interaction between lnx and dummy_APO_E . Both of them show the same t value with the same p value for proc reg and glm.

The code that I run looks like this:

proc reg data=merg3_data;
model AVAL =lnx interaktion dummy_APO_E ;
run;

proc glm data=merg3_data ;
class dummy_APO_E;
model aval = lnx |dummy_APO_E / solution;
run;

I am using SAS Enterprise Guide 8.3

Best regards,

Tobias

[PaigeMiller]

Probably, GLM has parameterized the model differently than you did in REG. This is normal, and nothing to worry about; it does have the possibility of different parameter estimates and t-values for the dummy variable and interaction. However, the model is the same, as indicated by your statement: "The overall F-test is the same for both proc reg and glm."