Hi,
I have a doubly multivariate data set in which for 18 subjects 2 measures (a discriminability measure and a physiological one) are taken on four occasions. I am trying to best come up with a measure of the correlation of the two measures across the 4 times taking into account the within subjects aspect of the data. A colleague of mine simply dumped all of 72 measurement pairs together and computed the overall correlation. This can not be the best way (even though it got past peer review). Running MANOVA (syntax below) I get reliable linear trends for each measure, which is nice. Is it sufficient to point to the high partial eta squared values for each contrast and say, in effect, voila?
Thanks in advance,
MB
MANOVA par_300_500_T1 to par_300_500_T4, pr_T1 to pr_T4
/wsfactors testnum(4)
/MEASURE = amp, pr
/TRANSFORM(par_300_500_T1 TO pr_T4 ) = POLYNOMIAL
/RENAME = AMP AMP_LIN AMP_QUAD AMP_CUB
PR PR_LIN PR_QUAD_ PR_CUB
/WSDESIGN = testnum
/PRINT = SIGNIF(UNIV, STEPDOWN, EFSIZE, GG) ERROR(CORR) HOMOGENEITY(BOXM)
/DESIGN.
**********************************************************
Michael Bersick
Assistant Research Scientist
[hidden email]
Cognitive Electrophysiology Laboratory
New York State Psychiatric Institute
1051 Riverside Drive, Unit 6
New York City, NY 10032
(212) 543-6626
Harvesting brain grain for a better tomorrow...
**********************************************************
Locked
