topic Re: MANOVA intercept in Statistical Procedures

MANOVA intercept

TerrieShaft — Mon, 10 Oct 2016 15:41:20 GMT

I'm running a MANOVA and I get identical f-values when I ask for the intercept and when I omit it from my statement. This seemed counter intuitive to me, and also to my colleague who uses MANOVA extensively (however he uses SPSS). Can anyone help me understand why this makes no difference? I've included my code below. I'm sure I'm overlooking something basic.

proc glm;
class delay familarity breadth;
model attitude perfave = delay familarity breadth delay*familarity delay*breadth familarity*breadth
delay*familarity*breadth;
manova h= intercept delay familarity breadth delay*familarity delay*breadth familarity*breadth delay*familarity*breadth /summary;
manova h= delay familarity breadth delay*familarity delay*breadth familarity*breadth delay*familarity*breadth /summary;
run;

Re: MANOVA intercept

Rick_SAS — Mon, 10 Oct 2016 16:57:02 GMT

There are many F values printed, so I'm not sure which ones you are concerned about. Are you talking about the MultStat table?

Run the following program and look in the log to find the name of the tables:

proc glm data=sashelp.class;
class sex age;
model weight height = sex | age;
manova h= intercept sex|age /summary;
manova h= sex|age /summary;
run;

Re: MANOVA intercept

Ksharp — Tue, 11 Oct 2016 12:46:14 GMT

model weight height = sex | age/ nouni;

Re: MANOVA intercept

TerrieShaft — Tue, 11 Oct 2016 21:39:09 GMT

Thanks. I think I finally sorted it out. SAS fits an intercept b/c of the class variables - I didn't realize it until I tried to specify NOINT (see SAS log details below). I suspected it was including an intercept but this verifies it.

NOTE: Due to the presence of CLASS variables, an intercept is implicitly fitted. R-Square has
been corrected for the mean.