 |
 |
Thanks Donald. So, (if I get it correctly) your suggestion is to estimate betas for each condition (5 in total) and then compute pairwise contrasts such as [ 1 -1 0 0 0], [0 1 -1 0 0], [0 0 1 -1 0] and [0 0 0 1 -1]. In this way I obtain 4 maps for each participant that I can submit to 2nd level analysis (t-tests). Then, the last step would be to use these four maps to create a sort of conjunction map, revealing voxels where a significant increase between each pair overlaps, is it correct? Do you have any suggestions about the criteria (e.g. p-threshold and cluster size) that can be recommended for this analysis, particularly when I do the conjunction map? Thanks, D 2015-04-06 17:26 GMT+02:00 MCLAREN, Donald <[log in to unmask]>: > To be clear, the linear contrast (-2,-1,0,1,2) does not mean that each > condition must be higher than the previous condition. You are testing > whether the linear contrast*beta is different than 0. This is true whether > it is a contrast or parametric modulator - although you will get different > values because one is built into the model and one is not built into the > model. > > Based on your question, you want to know which regions have significant > increases from condition to condition. For this, I would model each > condition separately, then compute the difference between each pair-wise > condition comparison, then use them in one-sample t-tests to identify > voxels that increase between each pair. Finally, I'd look at where those > voxels overlap between the 5 pair-wise comparisons. > > Best Regards, Donald McLaren > ================= > D.G. McLaren, Ph.D. > Research Fellow, Department of Neurology, Massachusetts General Hospital > and > Harvard Medical School > Postdoctoral Research Fellow, GRECC, Bedford VA > Website: http://www.martinos.org/~mclaren > Office: (773) 406-2464 > ===================== > This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED > HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is > intended only for the use of the individual or entity named above. If the > reader of the e-mail is not the intended recipient or the employee or agent > responsible for delivering it to the intended recipient, you are hereby > notified that you are in possession of confidential and privileged > information. Any unauthorized use, disclosure, copying or the taking of any > action in reliance on the contents of this information is strictly > prohibited and may be unlawful. If you have received this e-mail > unintentionally, please immediately notify the sender via telephone at > (773) > 406-2464 or email. > > On Sat, Apr 4, 2015 at 1:01 PM, David <[log in to unmask]> wrote: > >> Dear all, i would like to get some advice on the following procedure. >> I performed a GLM with two parametric modulators (PM) from an event >> related design. >> In the design there is a single event of interest and two modulators that >> model the event-related activity as a function of five conditions. >> The two PMs have the form: >> PM(1) (Conditions from 1 to 5) = - 3,-3,2,2,2 >> PM(2) (Conditions from 1 to 5) = - 2,-1,0,1,2 >> >> So, one resembles a step function and the other is a linear increase. The >> reason why I used parametric modulation is because I want to identify >> regions where my experimental conditions induce a linear (and not stepwise) >> increase in activity. Being these two patterns quite similar, I thought >> parametric modulation analysis in spm would be the optimal strategy, since >> I can get betas for the linear PM (PM(2)) after "removing" all the variance >> explained by PM(1). >> >> However, when I inspect locally the pattern of activity as a function of >> the five conditions in clusters that result significant for the PM(2), I >> observe that in some clusters the pattern is different from the expected >> linear increase (e.g., 1 4 -3 0 2 or similar, just to make an example). >> Now, unless I am doing something wrong, I think this could be due to the >> orthogonalization between sequential modulators applied by SPM. >> So, what could be the solution to get a statistical map with exclusively >> (if any) significant clusters with linear (and not stepwise) increase? >> >> I was thinking to do a conjunction analysis with the statistical map >> obtained for the PM(2) and a map obtained with a separate GLM in which the >> five conditions are regressed separately and then contrasted with the >> specific contrast (-2,-1,0,1,2). This should at least results in a >> statistical map where significant voxels (if any) are showing activity that >> is not explained by the mere stepwise function (PM(1)) and that resembles >> exclusively a linear increase (P(2)*Linear_contrast)… Could this be a >> reliable solution? >> >> Since I am quite new in fmri analysis, any advice will be very >> appreciated. >> Thanks >> > > -- Pascucci David __________________________________________ Department of Neurological and Movement Sciences Section of Physiology and Psychology, University of Verona Strada Le Grazie 8, I-37134 Verona, Italy