 |
 |
nope, the cope is the group mean. The zstat is the transformed tstat and you want the numerator of the tstat, which is the mean. This is the cope tstat=cope/sqrt(varcope) zstat=t_to_z_transform(tstat) Cheers, Jeanette On Mon, Feb 4, 2013 at 1:36 PM, Tseng Mark <[log in to unmask]> wrote: > So, just for confirmation: Is the t (zstat) map in my higher-level > analysis (1-sample t test) exactly the group mean of this 3-way interaction > effect? > > Thanks again. > > 2013/2/4 Jeanette Mumford <[log in to unmask]> > >> Hi, >> >> Yes, the thresh_zstat image you found is the thresholded version of the >> stats/zstat image. This zstat image is what results after the t-to-z >> transform is applied to the tstat image. This is done so you don't need to >> worry about differences in degrees of freedom (zstat maps are comparable >> across studies, tstat maps only are if the DF match). The tstat map is the >> ratio of the cope divided by the square root of the varcope file. >> Basically the cope file is the numerator of your t-test statistic, which >> in this case is the mean (I'm assuming your group model was a column of 1s). >> >> You will have a cope.feat directory for every lower level *t-test. *F-tests >> cannot be brought to higher levels. So, if you have 8 contrasts specified >> in your first level analysis, you'll have 8 cope.feat directories in the >> next level (if you applied your model to all lower level copes). >> >> I hope that helps. Again, nothing about your first level f-tests will be >> transferred to higher levels. >> >> Jeanette >> >> On Mon, Feb 4, 2013 at 12:14 PM, Tseng Mark <[log in to unmask]> wrote: >> >>> Sorry Jeanette, my stupid further ask..... where is the cope estimate? >>> >>> For example, in my first-level model, my f contrast is just using >>> contrast 3. After higher-level analysis for group mean (1-sample t tests of >>> many first-level contrasts), in the cope3.feat folder, there is a file >>> called thresh_zstat.nii.gz (this should be the thresholded t map) as well >>> as a stats folder containing a file called cope.nii.gz. Is this what you >>> mean? I don't think so because this file also exists in other cope.feat >>> that has no f contrast. >>> >>> 2013/2/4 Jeanette Mumford <[log in to unmask]> >>> >>>> Hi, >>>> >>>> The group t-stat map will be a test of that effect and corresponds to a >>>> 1-sided test. The actual mean is not the t-stat, but the cope estimate. >>>> >>>> Cheers, >>>> Jeanette >>>> >>>> >>>> On Mon, Feb 4, 2013 at 8:38 AM, Tseng Mark <[log in to unmask]> wrote: >>>> >>>>> Hi Jeanette, >>>>> >>>>> Ah....this is actually a continuous question of my previous post "2x2x2 >>>>> repeated measures ANOVA model." >>>>> >>>>> Just like we discussed, presumably at the first level the order of the >>>>> 8 EVs is: >>>>> A1B1C1 >>>>> A1B1C2 >>>>> A1B2C1 >>>>> A1B2C2 >>>>> A2B1C1 >>>>> A2B1C2 >>>>> A2B2C1 >>>>> A2B2C2 >>>>> >>>>> The 3-way interaction in my first-level model would be: >>>>> ABC =[1 -1 -1 1 -1 1 1 -1] >>>>> >>>>> I just want to know the group mean of this 3-way interaction. >>>>> >>>>> By setting up a group mean 1s model in my higher level analysis, >>>>> should I just check the t (zstat) map which is exactly the group mean of >>>>> this 3-way interaction? >>>>> >>>>> Thanks in advance. >>>>> >>>>> Mark >>>>> >>>>> >>>>> 2013/2/4 Jeanette Mumford <[log in to unmask]> >>>>> >>>>>> Hi, >>>>>> >>>>>> The tricky thing with an f-contrast is that there isn't a single mean >>>>>> as you are testing a group of contrasts simultaneously in a "are any of >>>>>> these significant?" test. In FSL you cannot bring lower level f-contrasts >>>>>> to the higher level, but will work with single contrasts instead. >>>>>> >>>>>> Hope that helps, >>>>>> Jeanette >>>>>> >>>>>> >>>>>> On Mon, Feb 4, 2013 at 8:18 AM, Mark <[log in to unmask]> wrote: >>>>>> >>>>>>> Hi, >>>>>>> >>>>>>> Sorry this might be a simple question but I just don't know how..... >>>>>>> >>>>>>> In my first-level feat model, I have a F contrast as well as many t >>>>>>> contrast. I want to see the group mean of the F contrast. However, when I >>>>>>> set up a group mean contrast using 1s in the higher-level model, I can only >>>>>>> see the group average of t contrasts. >>>>>>> >>>>>>> How can I see group mean of F contrast? >>>>>>> >>>>>>> Thanks. >>>>>>> >>>>>>> Mark >>>>>>> >>>>>> >>>>>> >>>>> >>>> >>> >> >