Mr Jigar Jogia wrote:
> I have a set of data with patients and control subjects i have
> preprocessed the data using spm99 (co-reg, normalised, seg & smoothed)
>
>
> With regards to the analysis i want to look for differences between
> patients and controls, with gender (male/female) as another independent
> factor and age as covariate.
>
> The next step i believe is 3. Model specification & parameter estimation
> i believe the model to analyse this imaging data is Multi-subject,
> conditions & covariates
>
> however i am not sure on the actual steps i need to take to carry out this
> analysis.
>
Typically this sort of analysis would involve two stages of modelling:
1. Modelling individual subjects.
2. Modelling average effects in each group.
sometimes referred to as the 1st and 2nd levels. This separation
of levels assumes that you
wish to make inferences about the populations from which the subjects
are drawn (so-called 'random effects analysis'). This takes place
at the second level. Your 2nd level would involve a two-sample t-test
at each voxel which would compare the average effects for
patients versus controls. Actually your 2nd stage model would be more
complex as you'd need to have age and sex as regressors as well.
For more on the 2nd level see eg. first 7 slides of
http://www.fil.ion.ucl.ac.uk/~wpenny/notes03/slides/hier/slide1.htm
To model individual subjects you need to set up a separate GLM and
SPM analyses for each subject according to eg.:
http://www.fil.ion.ucl.ac.uk/~wpenny/notes03/slides/glm/slide1.htm
See also
http://www.fil.ion.ucl.ac.uk/~wpenny/notes03/slides/event/slide1.htm
for more advanced fMRI designs.
For more on these issues see papers on our web page:
http://www.fil.ion.ucl.ac.uk/spm/course/notes.html.
The most up-to-date info is in our new book 'Human Brain Function II'
which you could persuade your local library/lab to buy (its 150ish pounds).
From each level 1 analysis you will
make a 'contrast' image. This captures or 'summarises' the effect
for each subject (the two-level approach is sometimes called the
summary statistic approach). You then enter these 'con' images as the
data for level-2.
Hope this helps.
Best, Will.
> Best Wishes
>
> JJ
>
>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7478
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|