Dear SPMers
We have a pre-/post-treatment SPECT analysis with 3 subject groups:
controls (2 scans each),
patients who responded to treatment (3 scans each: pre- and post-
treatment and long-term
follow-up) and patients who didn't respond (2 scans each: pre- and
post-treatment).
We have in addition a set of pre-treatment scans for other patients
whose response to treatment
is unkown for varying reasons.
We have currently set up a multi-group multi-condition model using only
controls and
patients of known response. This allows us, among other things, to
compare post-treatment
responders to post-treatment non-responders.
There is a debate whether to include the scans of the patients of
unknown response as a
separate, 4th group, since some members of our group suggest this would
increase the
statistical power of the analysis in the comparison of all patients'
pre-treatment scans to controls.
Please does anyone have an opinion as to whether this would be a
sensible thing to do?
Does the definition of a separate block for this group of subjects,
according to the GLM,
hypothesise that this group of patients has a unique pattern of
perfusion different to those
of the other 3 groups? Might it weaken the statistics of the other 2
patient groups?
The alternative is to retain the current model (3 groups only) and
additionally set up a separate
design containing only 2 blocks: 1st scans of controls, and 1st scans of
all patients (including
the mixed group).
With many thanks for any comments,
Hava
--
************************************************************************************
Hava Lester, PhD [log in to unmask] +972 (0)2 677 7148
home: [log in to unmask] +972 (0)2 567 0778
Dept Med Biophysics & Nucl Medicine, Hadassah Hospital, Ein Kerem, Jerusalem, Israel
|