Dr. Nichols et al,
I'm hoping you can help with a particular analysis
I'm trying with SPM. I'm sure I know just enough to
be dangerous, and as I follow the mailing list it
seems like you are the one who could most likely
answer my questions. And my good friend Mike Glabus
is losing patients (oops, I mean patience) with me...
I have a pilot study where two animals were imaged
4 times: once pre-lesion and 3 times post lesion
at (1,4, and 8 wks).
I have created a normalization template, coregistered,
normalized and smoothed the images.
My hypothesis is that PET imaging with FMT is useful
for seeing these lesions, and my first thought is that
a paired t-test, where each pair contains the pre-lesion
and one of the post lesion pairs, would be appropriate.
This gives 6 pairs.
Here's my confusion: According to Mike, I can't use
the "Basic Models" paired t-test, because a t-test is
a second level analysis, and also maybe because of
fixed vs. random effects.
1) Should I care about fixed/random? It seems to me
that I am not trying to make population inferences
here so a fixed analysis would be valid. In my
reasoning, the between subject variability is not
an issue if the pairs of the t-test are from the
same subject.
2) I have gleaned from the spm list that one way to
create the second level analysis would be to take
the image pairs and create one subtraction image
for each pair,
and then do a one sample t-test... how exactly is
this different from a paired t-test?
3) I have "normalized" the intensity of these images
by measuring non specific uptake in the cerebellar
region and dividing the entire image by the mean
value, and doing this for each image independently.
Do I have to worry about sphericity correction, and
if so, are replications over pairs or conditions?
4) If not a t-test, then what?
Hope you can help me understand a little better.
Thanks a bunch,
Kerrie Tainter
Kerrie Holton Tainter, PHD
Research Imaging Scientist
PET Imaging Center
Biomedical Research Foundation
Shreveport, LA
318-675-4038
[log in to unmask]
|