I'm running some analyses where SPSS and SAS are giving me different results (i.e., significant here, but not there). I'm attaching the output. I've mostly fiddled with the SAS syntax to match the SPSS results. I've tried adding the covariates to the class statement, removing the random effects line, making the random covariance structure AR(1), etc.
I THINK SAS is using only 3 parameters (gender, time, their interaction) while SPSS is using more parameters (2 levels of gender * 13 levels of time * 13 interaction terms). Here's the syntax:
SAS:
proc mixed covtest;
class id weekcat ;
model SOBFieldWD= weekstart0|female domestic semester/s;
random intercept /subject=id type=un;
repeated weekcat/subject=id type=ar(1);
run;
SPSS:
MIXED
sobfieldwd by female weekstart0 with domestic semester
/FIXED = female weekstart0 female*weekstart0 | SSTYPE(3)
/METHOD = REML
/PRINT = R SOLUTION TESTCOV
/REPEATED = weekstart0 | SUBJECT(ID) COVTYPE(ar1) .
I THINK SAS is using only 3 parameters (gender, time, their interaction) while SPSS is using more parameters (2 levels of gender * 13 levels of time * 13 interaction terms). Here's the syntax:
SAS:
proc mixed covtest;
class id weekcat ;
model SOBFieldWD= weekstart0|female domestic semester/s;
random intercept /subject=id type=un;
repeated weekcat/subject=id type=ar(1);
run;
SPSS:
MIXED
sobfieldwd by female weekstart0 with domestic semester
/FIXED = female weekstart0 female*weekstart0 | SSTYPE(3)
/METHOD = REML
/PRINT = R SOLUTION TESTCOV
/REPEATED = weekstart0 | SUBJECT(ID) COVTYPE(ar1) .