 |
 |
Hi Anderson, It is a randomised placebo controlled design. We have a group of patients assigned to either active treatment or placebo. We want to asses the effect of treatment vs. placebo. One can say it is from baseline to follow up while controlling for the placebo effect. By the book, this would just be the interaction effect baseline and follow-up vs placebo and active. However, since we know this is not a sensitive design for fMRI data due to high within subject variability, the idea is that we assume no difference between subjects at baseline and then just compare the active and placebo groups at follow-up. But it would be nicer to be able to add the baseline data as covariates since the assumption that they are the same at baseline is not waterproof. The latter approach would be justified if we are able to see that baseline and follow-up scans are indeed correlated. Best, Julian On Sun, Jun 16, 2019 at 12:43 AM Anderson M. Winkler <[log in to unmask]> wrote: > Hi Julian, > > I'm sorry for the late reply. Was attending conferences. Now I realise > there is at least one more piece of information missing. What are the > hypotheses for your two studies? Are you interested in differences between > baseline and follow-up, or are these two scans intended to improve ability > to detect effects that would persist across both timepoints? > > All the best, > > Anderson > > > On Wed, 12 Jun 2019 at 08:52, Julian Macoveanu <[log in to unmask]> > wrote: > >> Hi guys, >> >> I would still need some input on the issue below. >> >> All the best, >> Julian >> >> On Mon, May 27, 2019 at 10:08 AM Julian Macoveanu < >> [log in to unmask]> wrote: >> >>> Hi Anderson, >>> >>> I have not done the correlation yet, how to do it was actually my first >>> question. The only way I see it, but please correct me if I am wrong, is to >>> look at one contrast at a time, e.g. I enter the cope1 from one group as >>> input, and add an voxel-wise EV in the stats which is a 4D made from cope1s >>> of the respective group at follow-up (in matching order as the input). I >>> would then look at the contrast to check for any significant effects (+/- >>> 1). >>> >>> Regarding the study design, there are actually two studies we planned to >>> use this approach on: >>> - Randomised trial with patients assigned to either real treatment or >>> placebo >>> - Remitted patients that that had a secondary episode vs. remitted >>> patients that did not have a secondary episode after one year. So all >>> patients scanned at the first remission, and then again after one year. >>> >>> Best, >>> Julian >>> >>> >>> >>> On Sat, May 25, 2019 at 11:34 PM Anderson M. Winkler < >>> [log in to unmask]> wrote: >>> >>>> Hi Julian, >>>> >>>> Could you tell the following: >>>> >>>> - How are the correlations between baseline and follow-up computed? Do >>>> you do some kind of spatial correlation between results, or do you check >>>> whether task-based BOLD-responses are correlated in both scans? Or >>>> something else? >>>> - How are the groups defined? Is this a randomized controlled trial or >>>> just an observational study of the kind patients vs. controls? >>>> >>>> All the best, >>>> >>>> Anderson >>>> >>>> >>>> On Fri, 24 May 2019 at 12:50, Julian Macoveanu < >>>> [log in to unmask]> wrote: >>>> >>>>> Hi all, >>>>> >>>>> I have 2 groups scanned with fMRI twice. Due to the known high within >>>>> subject variability which approaches between-subject values, my group >>>>> decided to first perform a correlation test between baseline and follow-up >>>>> measurements. If there is a significant correlation between baseline & >>>>> follow-up then we would like to perform a 2-sample t-test using the >>>>> follow-up data while adjusting for baseline data. If the correlation test >>>>> is negative i.e. the baseline cannot predict the follow-up, then we would >>>>> just do the 2-sample follow-up t-test without baseline correction. Instead >>>>> we would simply check that whatever we find between groups at follow-up was >>>>> not present between groups at baseline. >>>>> >>>>> Now, the question is, do you think this is a sensible approach? Can >>>>> the correlation analysis be made in FEAT somehow using voxelwise >>>>> covariates? I guess if I find the answer to this question I can also move >>>>> on and similarly adjust the data by the baseline data as covariate. >>>>> >>>>> Julian >>>>> >>>>> ------------------------------ >>>>> >>>>> To unsubscribe from the FSL list, click the following link: >>>>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 >>>>> >>>> >>>> ------------------------------ >>>> >>>> To unsubscribe from the FSL list, click the following link: >>>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 >>>> >>> >> ------------------------------ >> >> To unsubscribe from the FSL list, click the following link: >> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 >> > > ------------------------------ > > To unsubscribe from the FSL list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 > ######################################################################## To unsubscribe from the FSL list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1