Not applicable
Posts: 1

# Proc GLM

Hello everybody,

I am trying to make an analysis of variance with repeated measures. I want to reconstruct an example of a book where the author says he got the results from SAS PROC GLM (but he gave no syntax).
This is my syntax:

data orig_daten;
input student\$ y1 y2 y3 y4 y5 y6 y7 y8 y9 y10;
cards;
1 7.79 8.01 7.79 8.37 8.12 8.08 7.86 7.64 8.23 8.05
2 7.33 7.33 7.64 7.71 7.57 7.01 7.48 7.33 7.33 7.30
3 7.86 7.94 8.08 8.08 8.02 8.08 8.08 8.08 8.37 8.23
4 7.48 7.79 7.71 7.56 7.61 7.71 7.79 7.79 8.01 7.89
5 7.56 7.64 7.64 7.94 7.78 7.64 7.86 7.79 7.48 7.63
6 8.01 7.71 5.96 7.79 7.54 6.50 6.41 5.96 6.24 6.27
7 7.33 7.33 6.92 7.94 7.57 6.92 6.92 7.71 7.71 7.46

;
run;

proc glm data=orig_daten;
class student;
model y1-y10=student / nouni;
repeated course 2 (10 20), period 5 (1 2 3 4 5) / printe;
run;

So there are two within-subject effects, course and period and one "between-subjects effect", the student.
When I run this programm I get the error, that no multivariate tests are performed due to insufficient error degrees of freedom.
This is because with this model I have got a three-way design and just one observation per cell. Because of this there aren't enough degrees of freedom left for the estimation of the error.
But the theory says, that one has to take the three-way interaction as the error term and calculate the anova with this Error.
But how can I tell SAS to take the three-way-interaction as error term? Or what have I to do to get results?
(I get the right sums of squares but no F- or p-values.
I would be very thankful if someone could help me!
Best regards,
anna
Discussion stats
• 0 replies
• 178 views
• 0 likes
• 1 in conversation