Hi there,
My study was a RCT with an immediate treatment group and a waitlist
control group assessed at 3 time points - pre-treatment, post-
treatment and at a 3 month followup. I have both intermittent
(occassional missing data due to incorrect questionnaire or subscale
completion) and missing data due to dropout. I have greater dropout
from my waitlist control group - a problem that is not so uncommon
from what I understand. I have small n's excluding my missing data -
immediate treatment group n=20 and waitlist control group n=10. I had
3 dropout of my treatment group (2 due to family deaths) and 14
dropout of my waitlist control group.
My problem is really modelling the missingness to determine whether it
is MCAR (probably not), MAR or MNAR. SPSS in missing value analysis
has Little's MCAR test for determining whether data are MCAR and
Separate Variance t-tests for determining whether data are MAR. My
problem is that I am unsure how to enter my variables to determine
these patterns of missingness. My study is one of psychology and we
tend to give questionnaires as measures. I have 6 questionnaires that
were repeated at times 1, 2, and 3. For most of these questionnaires
there are various subscales which purportedly measure slighlty
different things. My question is do I subject each subscale to the
MVA or do I combine all subscales from a particular measure for MVA. I
have tried both ways - getting a pooled Little's MCAR statistic when I
subject all subscales to MVA at once. I also tend to get significant
results when I use all subscales on the Separate Variance t-tests
compared to when I analyse each subscale separately.
I am also unsure whether I should be looking at each group separately
when it comes to analysing missing data patterns?
I also wonder whether my dataset is too small to be using MVA and
producing reliable results?
Unfortunately none of the examples I have read have variables similar
to mine so I am finding this part rather challenging so any help would
be gratefully appreciated.
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
