 |
 |
Hi Torben, On 09.01.2014 15:35, Torben Lund wrote: > > > > Den Uge:2 09/01/2014 kl. 14.55 skrev Paul Glad Mihai > <[log in to unmask] <mailto:[log in to unmask]>>: > >> Hi Torben, >> >> thanks for your quick reply. So the two data sets do indeed have unequal >> variances, which is good, because they don't violate the assumption of >> sphericity. However, I don't understand why both contrasts: >> >> [1 0] for young subjects >> [0 1] for old subjects >> >> result in identical contrasts with identical thresholds, > > The threshold depend on the df and the smoothnes of the residual image, > these numbers are the same for the two contrasts > >> while looking >> at them from an one sampled t-test perspectives gives different results. >> > > The one sample t-test would have a different df and smoothness, you > could try a [1 1] contrast in the two sample t-test this should give you > something similar the 1- sample t-test (use a [0.5 0.5] contrast if you > want the contrast image to be similar I used all images from both age groups in a one sampled t-test and the result is almost identical to the one from the two sampled t-test with contrast [1 1]. Even looking at the global maximum the coordinates and T values are the same. Thanks for the heads up! So basically the activations between the two groups are so similar, that the difference is negligible. This is a somewhat unexpected outcome, alas good news for older people :D. Regards, Glad > Hope this helps > Torben > > >> Any thoughts? >> >> Regards, >> Glad >> >> On 09.01.2014 12:59, Torben Lund wrote: >>> Dear Glad Mihai >>> >>> >>> The gray box indicates that you have specified that the two groups have >>> unequal variance. These variances is estimated by SPM and if they are >>> indeed different you will see different gray-levels, because the >>> designmatrix have now been whitened. You can also inspect the covariance >>> matrix directly, check the SPM manual for instructions on how to do this. >>> >>> >>> Best >>> Torben >>> >>> >>> Torben Ellegaard Lund >>> Associate Professor, PhD >>> Center of Functionally Integrative Neuroscience (CFIN) >>> Aarhus University >>> Aarhus University Hospital >>> Building 10G, 5th floor, room 31 >>> Noerrebrogade 44 >>> 8000 Aarhus C >>> Denmark >>> Phone: +45 7846 4380 >>> Fax: +45 7846 4400 >>> http://www.cfin.au.dk <http://www.cfin.au.dk/> >>> [log in to unmask] <mailto:[log in to unmask]> >>> <mailto:[log in to unmask]> >>> >>> >>> >>> >>> >>> Den Uge:2 09/01/2014 kl. 12.40 skrev Glad Mihai >>> <[log in to unmask] <mailto:[log in to unmask]> >>> <mailto:[log in to unmask]>>: >>> >>>> Hello SPM fans, >>>> >>>> I calculated a two sample t-test using data from young and old >>>> subjects and I got a strange result. >>>> >>>> If I look at the data for either the young or the old population in >>>> the two sample t-test scenario, I get almost no difference. >>>> If I calculate a one sample t-test for each group, I get a significant >>>> difference in activation with different T thresholds. Furthermore, >>>> when I compute the difference between the two in the two sample >>>> t-test, I get no significant activation, which should make sense, >>>> because there is no difference between the two contrasts inside the >>>> two sample t-test calculation. Additionally, the cutoff threshold for >>>> both is the same using the two sample t-test, which I find strange. >>>> >>>> I've attached some screen captures so you can see what I mean. >>>> >>>> The design matrix shows that for group 2 (older subjects) the box is >>>> gray. What is happening here? >>>> >>>> I did two measurements exactly two years apart using the same design. >>>> Is there a problem of independence? >>>> >>>> I also preprocessed them the same, using new segment and dartel >>>> normalization. >>>> >>>> Cheers, >>>> Glad >>>> <old_one_sample_ttest.png><old_two_sampled_ttest.png><young_one_sampled_ttest.png><young_two_sampled_ttest.png> >>> >> >> -- >> Glad MIHAI, M.Sc. Biomedical Physics >> Functional Imaging | University Clinic Greifswald >> Walther-Rathenau-Straße 46 | 17475 Greifswald | Germany >> Tel: +49 3834 86 69 44 | Fax: +49 3834 86 68 98 >> www.baltic-imaging-center.de <http://www.baltic-imaging-center.de> > -- Glad MIHAI, M.Sc. Biomedical Physics Functional Imaging | University Clinic Greifswald Walther-Rathenau-Straße 46 | 17475 Greifswald | Germany Tel: +49 3834 86 69 44 | Fax: +49 3834 86 68 98 www.baltic-imaging-center.de