Dear all,
This is a follow-up question on some posts from 2009 on missing values ("
missing values - randomise design", 13 Oct 2009).
In brief: I have a several behavioral variables that I want to correlate
with FA in TBSS using randomise. For each behavioral variable I have some
missing values or values that I identified as outliers and want to exclude
them. I would like to "exclude" some subjects from analysis without creating
a new 4D file for each variable.
I wanted to confirm if I understood the previous discussion:
If this is my design:
subjects beh missing1 missing2
sub1 beh1 0 0
sub2 0 1 0
sub3 beh3 0 0
sub4 beh4 0 0
sub5 beh5 0 0
sub6 0 0 1
sub7 beh7 0 0
sub8 beh8 0 0
sub9 beh9 0 0
With beh standing for different values of a behavioral measure.
My questions:
1) can i put "0" for missing values in beh column, or should I replace them
with group means?
2)If I want to check for correlation of beh data with FAwhat would be the
correct contrast?
3)In this particular case, can i replace 2 missing columns with just one
(and contrast 1 0)?
Thank you,
Cheers,
Aga
|