External Email - Use Caution
Hi Kersten,
I followed your mail and even also I got stuck at the contrast matrix i.e
CM:
I have given the code below which I used to execute the longitudinal
pipeline.
[aseg, asegrows,asegcols] = fast_ldtable(['aseg.long.table']);
asegcols=cellstr(asegcols);
[Y,mri] = fs_read_Y('rh.thickness_sm10.mgh');
rhsphere =
fs_read_surf('/usr/local/freesurfer/subjects/fsaverage/surf/rh.sphere');
rhcortex =
fs_read_label('/usr/local/freesurfer/subjects/fsaverage/label/rh.cortex.label');
Qdec = fReadQdec('long.qdec.table.dat');
Qdec = rmQdecCol(Qdec,1);
sID = Qdec(2:end,1);
Qdec = rmQdecCol(Qdec,1);
M = Qdec2num(Qdec);
[M,Y,ni] = sortData(M,1,Y,sID);
Y(:,all(Y==0))=[];
%lme_lowessPlot(M(:,1),Y(:,1)+Y(:,2),0.70,M(:,2));
X = [ones(length(M),1) M M(:,1).*M(:,2)];
[lhTh0,lhRe] = lme_mass_fit_EMinit(X,[1 2],Y,ni,[],3);
%[lhTh0,lhRe] = lme_mass_fit_EMinit(X,[1 2],Y,ni,[],5);
[lhRgs,lhRgMeans] = lme_mass_RgGrow(rhsphere,lhRe,lhTh0,[],2,95);
%[lhRgs,lhRgMeans] = lme_mass_RgGrow(lhsphere,lhRe,lhTh0,[],2,95);
lhstats = lme_mass_fit_Rgw(X,[1 2],Y,ni,lhTh0,lhRgs,rhsphere);
lhTh0_1RF = lme_mass_fit_EMinit(X,[1],Y,ni,rhcortex,3);
lhstats_1RF = lme_mass_fit_Rgw(X,[1],Y,ni,lhTh0_1RF,lhRgs,rhsphere);
LR_pval = lme_mass_LR(lhstats,lhstats_1RF,1);
CM = [0 1 0 0 0 0 0];
%%%%%%%%%%%%
F_lhstats = lme_mass_F(lhstats,CM);
My design matrix i.e X has 7 columns ['subjID(1)', 'time_years', 'Age',
'Years_of_education', 'Gender', 'ICV', 'time_yearsXAge' ]
So I kept the contrast vector as CM = [0 1 0 0 0 0 0] (if you remember my
mails, this study doesn't have a control group).
Matlab threw an error for this design and the same is highlighted below:
>
>
> *F_lhstats = lme_mass_F(lhstats,CM);Error using lme_mass_F (line 44)The
> number of elements (contrasts) in CM must be equal to the number of
> locations (length(stats))length(stats)= 149995 *
> *length(CM)= 7*
I know it may sound silly but I am really stuck at this point in time.
Would be glad if you could let me know where I am going off the path!
Thanks,
Vittal
On Fri, Dec 11, 2020 at 10:29 PM vittal korann <vittalkor...@gmail.com>
wrote:
> Thanks, Kersten.
>
> I am sorry, I didn't see your response in the FS blog.
> I will go through it once and let you know!
>
> Once again thanks a lot for your kind mail!
>
> With regards
> Vittal
>
>
>
> On Thu, Dec 10, 2020 at 6:35 PM Diers, Kersten /DZNE <
> kersten.di...@dzne.de> wrote:
>
>> Hello Vittal,
>>
>> I am (again) attaching my response to your message. This response has
>> been on the freesurfer list since Nov 30, maybe you did not see it?
>>
>> Best, Kersten
>>
>> *From:* Diers, Kersten /DZNE
>> *Sent:* Monday, November 30, 2020 6:27 PM
>> *To:* Freesurfer support list
>> *Subject:* Re: Longitudinal pipeline
>>
>> Hello VIttal,
>>
>> thanks for the info about the design.
>>
>> You wrote before that the Matlab variable X (i.e., the design matrix) has
>> a dimension of 47x6. X is a translation of the study design into a
>> numerical matrix, which is part of the statistical model. In your case, I
>> assume that he first column is a vector of ones (i.e. the
>> intercept/constant of the statistical model), and that the other columns
>> reflect the numerical data in the table that you sent, in the same order.
>> Correct me if I am wrong.
>>
>> Now, the contrast "matrix“ specifies the questions that you want to ask
>> to the data, i.e. the effect of a certain variable (or sometimes
>> combinations of variables) onto the outcome variable Y. I put "matrix" in
>> quotes, because in many cases a simple vector is sufficient; contrasts need
>> not necessarily be specified as a matrix, and contrasts specified as a
>> matrix are typically more complex than those specified by a vector.
>>
>> In any case, the contrast vector/matrix needs to have as many
>> elements/columns as there are columns in X, because the elements/columns
>> oft the contrast vector/matrix will be matched with (and therefore need to
>> correspond to) the columns of X. I.e. the nth element/column in your
>> contrast vector/matrix refers to the nth predictor variable in the
>> statistical model, i.e. the nth column of X. There are 6 columns in X in
>> your case.
>>
>> The entries in the contrast vector/matrix represent weights that you
>> assign to each variable in order to test for its effect. One often (but not
>> always or necessarily) uses +1 to test for a positive relation of the
>> chosen variable on Y, -1 for a negative effect, and zero for not
>> considering a variable for a given contrast. Therefore, the simplest
>> contrast is a vector with a single non-zero entry, and this already allows
>> for the testing of effects. In other cases, contrasts can have more than
>> one non-zero entry, or can be formulated as a matrix, but we don’t need
>> this at the moment.
>>
>> The major issue with your design in its current form is that you will not
>> be able to distinguish effects of time from effects of treatment, due to
>> the absence of a control group. This precludes, as far as I can see, a
>> clear interpretation of any changes in Y that you might observe. If you
>> still want to go ahead and test the time/treatment effect, just set the
>> corresponding element of the contrast vector to +1 or -1.
>>
>> Best,
>>
>> Kersten
>>
>> Am 10.12.2020 um 09:10 schrieb vittal korann <vittalkor...@gmail.com>:
>>
>> Hi Kersten
>>
>> Thank you for kind response.
>>
>> I think we need to talk about the design matrix X first (sorry, I did not
>> look at your xlsx file; I think it's probably better to send simple csv
>> files if at all).
>> I attached longitudinal data in csv format.
>>
>> Besides the covariates that you mention, the matrix needs to contain an
>> intercept, a variable for ?time?, a variable for ?group? (assuming you have
>> a control group also), and a group-by-time interaction term in order to
>> asses differential change over time (again, if you have a control group).
>> I do not have a control group. Right now I have only schizophrenia
>> patients who underwent 2 scans: first is baseline and second one is after 3
>> months of yoga therapy. Wish to know the how can I proceed with my availabe
>> data.
>>
>> Just to make sure: ?age? should be ?age at baseline? (i.e. constant
>> across time per subject), otherwise it will be confounded with time.
>> Yes, I used same across the study.
>>
>> Could you review (and possibly adapt) your design matrix X with respect
>> to which columns it contains, and write it in the reply?
>> I tried but couldn't get there.
>>
>> Any help would be greatly appreciated!
>>
>> Thanks,
>> Vittal
>>
>>
>>
>>
>>
>> <longitudinal_data (1).csv>
>>
>>
>> --
>> Kersten Diers
>> Image Analysis Group (AG Reuter)
>> German Center for Neurodegenerative Diseases (DZNE)
>> B.1.114 | Building 99 | Venusberg-Campus 1 | 53127 Bonn | Germany
>> https://secure-web.cisco.com/1QEbnXQmS5z7LkqTPYTOn3q8tRJjlV7vi2P1a882OfirZspsbiYPIXn8GmkFqEa80Nmr4a62SzuJZgAlB8brmvdIHrcrHVaKt4A8PO5ebQbZEugej_SXZQ5iYAqKfJFv-QrIbiVvgT-wGP6OXHCzNdWplr4CdZ6nTJ1HMnJt49iwko9bWAwU9KwslrxwCtQbZI8Rah3P4d-KBFoKjNkDG6ciRlEAwYKIqOBQWJ3FKJ7j0zQ9x-uHkq0bZOIZgCp5mGAUyO_Q-G9H2x-5m7ijUmQ/https%3A%2F%2Fwww.dzne.de%2Fen%2Fsites%2Fbonn%2Fresearch-groups%2Freuter.html
>> Phone: +49 / 228 / 43302 - 381
>>
>>
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
