How to do projection of SSFs into new group data?

[annib]

Hi Chris and gang,
Now that I've got my SSFs that distinguish my groups how do I now use
this data to test how well it can predict group membership in a new
data set.

thanks
Anna

[Christian Habeck]

Hi anna,

this is explained in the documentation:

Example_Exo_pattern_and_VAF_example.pdf


You can also do this by hand from the Matlab command line and people more versed with Matlab could bypass the idiot proof, but more clumsy, software. I could show you this and save a Matlab diary file if you want.


Chris

-- 
Christian Habeck
ch...@columbia.edu
http://www.cumc.columbia.edu/dept/sergievsky/cnd/habeck.html

[annib]

Thanks for that.

Had a look at instructions and fell at the first hurdle :)

Just to check: my original design was 3 groups of grey matter maps and
I found the best combination of PCs that distinguished these groups.,
now I want to see if this pattern will distinguish between these 3
groups in my new dataset of GM maps.

So When it says number of conditions excluding baseline answer:1
I should just put 0, as I only have one GM map per new subject.

When it says upload images for baseline condition is that my list of
new GM images?

When it says upload paths to pattern image is that the image created
by my previous PCA, Akaike Rank 1 on my original GM images not the
resampled Z image produced by the permutations.

And the output of selection 14 will give me just one column associatd
with the 1 pattern image I loaded previously, telling me the VAF in my
new group?

Then would I have to extract the new TPRs for each subject and see if
they reliably distinguish between the 3 gentoype groups using ANOVA or
do I do a Z-score for each of the new subjects compared to the SSFs of
the orginal GM maps of the middle group?

Thanks loads
Anna

On Mar 30, 4:15 pm, Christian Habeck <chab...@gmail.com> wrote:
> Hi anna,
>
> this is explained in the documentation:
>

> *Example_Exo_pattern_and_VAF_example.pdf
>
> *You can also do this by hand from the Matlab command line and people

> more versed with Matlab could bypass the idiot proof, but more clumsy,
> software. I could show you this and save a Matlab diary file if you want.
>
> Chris
>

> *

[Christian Habeck]

The resulting expression values can be contrasted with t tests or
anova. Or you can correlate the dependent variable of allele status
with the expression vector.

--
Sent from my mobile device