 |
 |
On Mon, 21 May 2012 00:46:26 -0400, "MCLAREN, Donald" <[log in to unmask]> wrote: > On Mon, May 21, 2012 at 12:05 AM, Maryam Ziaei < > [log in to unmask]> wrote: > >> Dear SPM ers; >> >> I have some questions regarding the confusion due to reading so many >> responses in the archives. >> >> I have 5 conditions in my study with two groups of old and young people. >> Thanks God that I'm not interested in comparing all conditions, therefore >> in the first level I built a contrast let's say condition 2-baseline (0 >> 1 0 >> 0 -1)=A and condition 4 - baseline (0 0 0 1 -1)=B >> Then in group level I used these contrasts to compare between young and >> old in full factorial design, because I want to see the interaction. So, >> here are my questions: >> >> 1. Is the way of defining my first level contrasts correct or should I >> use >> Imcalc? the results are different. >> > > They are fine. > > >> >> 2. why in second level analysis some people build contrast like 1 0, the >> sum of contrast should be equal to zero, does it depend to the hypothesis >> or it is a rule? >> > > It depends on the null hypothesis. > > > >> >> 3. in full factorial condition, I'm interested to see if old group showed >> greater activation than young group, should I use t-test instead of >> F-test? >> the results are different as well (as one would expect). >> > > Assuming that you only used one contrast image per subject, you will have > two columns. Since you have a directional hypothesis, you should use a > t-test. An f-test will show you where either group is greater than the > other. The inference from an f-test is 2-sided; whereas in SPM, the > inference from the t-test is one-sided. > > >> >> 4. IF I wanna use the contrasts A and B to see where are the common areas >> in those condition across groups, should I use conjunction analysis, if >> yes, if conjunction analysis is the same as using inclusive mask? >> > > I usually create 2 maps. One for contrast A and one for contrast B. Once > you save the thresholded maps; use imcalc to compute the overlap of the > significant voxels (i1>0+2*(i2>0)). The overlap will have values of 3. > > >> >> The responses would be appreciated alot >> -- >> ********************** >> Maryam Ziaei >> PhD Student >> Department of Psychology (Cognitive Science) >> Stockholm University >> Tel: +46-8-163845 >> Fax: +46-8-159342 >> Alternative Email: [log in to unmask] >> Thanks alot for the responses; To make sure I understood it correctly, I made map with con images for each group separately for the interested contrast at 0.0001, then use the functions that you mentioned and then use the output in the group level analysis as an explicit mask. is that the way it should be done? If I do this can I write in my paper I use conjunction analysis or I should use the term of mask? To follow up, what are the differences between the way that you explained for the conjunction analysis and the way of selecting different contrasts and selecting conjunction analysis while you want to define contrast (as it asks for 'null hypo. to assess? conjunction, global and intermed ) with the way that you suggest? Thanks again for your helpful responses -- ********************** Maryam Ziaei PhD Student Department of Psychology (Cognitive Science) Stockholm University Tel: +46-8-163845 Fax: +46-8-159342 Alternative Email: [log in to unmask]