 |
 |
Hi Gwenaelle, Thank you for your reply! That is very helpful. Can I also clarify. What does it mean therefore when in the forums, a demeaned covariate is added but are not included in the contrasts? So for example, if you have 2 groups (3 participants in each) and 1 behavioural variable. Group EV1 EV2 EV3(behavioural variable, demeaned) 1 1 0 34.24 1 1 0 37.58 1 1 0 34.24 2 0 1 -12.42 2 0 1 -19.09 2 0 1 -49.09 Contrast1: 1 -1 0 Contrast2: -1 1 0 What do these contrasts look at? Does it ignore the demeaned covariate or do the contrast that these into account when comparing the two groups? Is this only done for covariates of non-interest? Regards, Sharpley On 30/04/2010 4:54 AM, Gwenaëlle DOUAUD wrote: > Hi Sharpley, > > My questions are: 1. Do I need to demean covariates if it is a > covariate of interest? 2. How do I set up the design matrix? Do I only > enter in the covariate values for the patient group? > > -> I guess it depends on your question. If this is a covariate of > interest, the easiest is to enter it in a simple design.mat with one > EV (say you've got three patients with 60, 70 and 80%): > 60 > 70 > 80 > > contrasts: 1 (if you expect a positive correlation) > and -1 > > Now if you want to investigate the difference in correlations between > patients and controls, you need to EVs (say you've got two controls > with 90 and 96%): > 90 0 > 96 0 > 0 60 > 0 70 > 0 80 > > contrasts: 1 -1 > and -1 1 > > > Hope this answers your question, > Gwenaelle > > > --- En date de : *Mar 27.4.10, S. Hsieh /<[log in to unmask]>/* a > écrit : > > > De: S. Hsieh <[log in to unmask]> > Objet: [FSL] covariate of interest in VBM > À: [log in to unmask] > Date: Mardi 27 avril 2010, 8h44 > > Dear FSL experts, I am a novice at using VBM and I need help with > correlating a behavioural variable of interest with grey matter > volume loss. I have read the threads in the forum but I haven't > been very successful at finding what I need at the moment > unfortunately. Let's say I have 2 groups (patient, controls) and > n=10 in each group. Both these groups also have a score on a > behavioural measure. Let's say this is percentage correct score. > What I am interested is the areas of atrophy that correlates with > this behavioural measure for the patient group only. My questions > are: 1. Do I need to demean covariates if it is a covariate of > interest? 2. How do I set up the design matrix? Do I only enter in > the covariate values for the patient group? > > > > For example: Group Patients Controls Covariate 1 1 0 80 1 1 0 70 1 > 1 0 20 1 1 0 66 ... 2 0 1 0 2 0 1 0 2 0 1 0 2 0 1 0 3. How do I > set up the contrast? If I have put the covariate only for the > patient group, i.e. do I put -1 (and the rest are zeros)? For > example: C1 0 0 -1 I'm not sure if I have explained myself > properly. Please let me know if you need any more information. > Thank you in advance for your help! Sharpley > --------------------------------------------------------------------------- > Sharpley Hsieh FRONTIER - Frontotemporal Dementia Research Group > Prince of Wales Medical Research Institute (POWMRI) Cnr Barker St > and Easy St Randwick NSW 2031 Australia Post: PO Box 82, St Pauls, > NSW, 2031 Phone: + 61 2 9399 1636 Fax: + 61 2 9399 1047 Email: > [log in to unmask] Website: http://www.ftdrg.org/ > > -- ---------------------------------------------- Sharpley Hsieh FRONTIER - Frontotemporal Dementia Research Group Prince of Wales Medical Research Institute (POWMRI) Cnr Barker St and Easy St Randwick NSW 2031 Australia Post: PO Box 82, St Pauls, NSW, 2031 Phone: + 61 2 9399 1636 Fax: + 61 2 9399 1047 Email:[log in to unmask] Website:http://www.ftdrg.org/ ----------------------------------------------------------------------- This email, and any attachments, may be confidential and also privileged. If you are not the intended recipient, please notify the sender and delete all copies of this transmission along with any attachments immediately. You should not copy or use it for any purpose, nor disclose its contents to any other person. -----------------------------------------------------------------------