James,
I suspect that you have not constructed your design matrix correctly. It is
difficult to comment in detail without knowing more about the experiment
and the ways in which you intended to analyse the data. However, the
following points might help.
Your design matrix has specified a linear increase over scans 1 through to
7 only. Scans 8 to 74 appear to have been weighted with a constantly high
value. If you look at your plots, this is what is depicted in plot 2. Also,
the linear increase you are looking for should span scans 1 to 7 according
your description. However, your model shows that the change spans across
scans 1 to 8. I am not sure what columns 8-10 represent.
If I understand you correctly, you are looking for areas in which changes
in perfusion are greater in the treatment group compared with the control
group. I assume that the groups have each been treated in exactly the same
way. You are therefore looking for differences between groups in terms of
the way that their values change over time.
I would suggest the following method of analysis. In a fixed effects
design, for a given subject, each of the 7 scans should occupy its own row
and its own column (as in cols 1-7 of your existing design matrix). The
first scan for other subjects would occupy the same column as scan 1 of
subject 1, but a different row (row 8 for subject 2, row 16 for subject 3
etc). Scan 2 for subject 1 would occupy row 2 of column 2 etc. The other
group would have exactly the same arrangement, except that the scans would
occupy an adjacent set of columns and a set of rows below those of the
first group. Your confounding covariates would then occupy the columns
after this.
The first 7 columns would represent scans 1-7 of your experimental group,
and columns 8-14 would represent scans 1-7 of your control group. In order
to look for an interaction, use a model that specifies changes going in
opposite directions in each group (e.g. ascending in experimental,
descending in control). Alternatively, I would specify no change in the
control group and an increase in the experimental group (ensuring that the
mean of the contrasts is zero). I would also use the SPM{F} map to specify
areas in which there was either a relative increase or a relative decrease.
You might also want to think about the following issue. One could only
infer that the effects are specific to the task if the non-specific effects
are sampled in a baseline condition. A time x condition x group interaction
would have allowed you to infer that your effects were task, as well as
group specific. If you do not have an independent baseline, then your
current design will only allow you to infer that there were differences
between the groups. These may relate to a combination of specific factors
(i.e. treatment regime) and non-specific factors that are not controlled for.
Best wishes,
Narender Ramnani
At 11:32 AM 9/23/00 -0500, you wrote:
>List,
>
>I am trying to determine exactly what the difference is between an
>"adjusted" VOI value is- and one where no adjustment has taken place.
>See this gif for a graphical description:
>
>http://neuropet-bri.org/spm/spm01.gif
>
>I am looking for linear changes in perfusion with treatment over seven
>scans, as compared to a group of untreated controls. After I produce the
>Z-map, I placed a 5mm VOI over a blob. The two graphs at bottom represent
>adjusted and nonadjusted responses, respectively.
>
>I guess my questions are:
>
>1. How do I get the mean voxel value from the VOI after the image has been
>globally scaled? Is this what "no adjustment" means?
>
>2. Why do neither of these graphs (see link below) of the patient VOI values
>compared to the mean of the control VOI values bear any resemblance to the
>contrast that I am using? In other words, this does not look like linear
>changes to me, despite the fact that it was produced by the contrast shown.
>Can anyone explain this, please?
>
>Thanks,
>
>James Patterson
*******************************
Dr Narender Ramnani
Sensorimotor Control Group
Department of Physiology
University of Oxford
Parks Road
Oxford OX1 3TP
Oxford University Centre for
Functional Magnetic Resonance Imaging of the Brain,
John Radcliffe Hospital,
Headington,
Oxford OX3 9DU
Tel. 01865 222704 (Direct)
01865 222729 (Admin)
mob. 0771 2632785
Fax. 01865 222717
email [log in to unmask]
********************************
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|