 |
 |
Hi, this is simply because the t-test is one-tailed and the F-test is two-tailed. I.e. for the t-test you accept a false-positive rate of 0.05 or less under one tail of the distribution whereas in the F-test at 0.05 you may accept only 2.5% for each tail. So alternatively, you may lower the threshold for the t-test to 0.025. Cheers, Andreas Am 07.10.11 08:52 schrieb "Chou Kun Hsien" unter <[log in to unmask]>: >Hello Andreas > >thanks for your kindly response >i've tried the threshold at p<0.1 as u said,and... it did work!! >we got a better result as we expected. >but we are just wondering why? how could we use p<0.1 rather than p<0.05? >would u please explain the reason for us? > >thank u so much, >Chou Kun Hsien > >>Hi, >>have you looked at p<0.1 instead of 0.05 for the (two tailed) F-test, >>does >>that match better with your t-tests? >>Cheers, >>Andreas > > > > > > >>Am 07.10.11 06:39 schrieb "Chou Kun Hsien" unter >><[log in to unmask]>: > >>>Sorry that i didn't clarify my question. >>>the following is the original question asked by Dr. John Colby >>> >>>>What I am confused about is this: When age is included as a covariate >>>>(as >>>>opposed to when it is not), the individual FA_tfce_corrp_tstat maps are >>>>more >>>>robust, but the overall F-test map (FA_tfce_corrp_fstat) drops almost >>>>to >>>>zero. This opposite behavior surprised me and makes me think that I >>>>might be >>>>doing something wrong in my model setup, or perhaps just not fully >>>>understanding how F-tests work in FSL. Does this seem like normal >>>>behavior, >>>>or does it look like I'm doing something wrong? >>> >>>we also met the same question like above. >>>the f-test result seems a lil bit different with the t-test result from >>>each t-contrast. >>>the f-values were small, and we can barely detect any significant >>>difference from f-test in TFCE (while p<0.05). >>>i think it is not a normal result. >>>cuz the f-test findings should not be so different from t-test, since we >>>got many significant findings from t-test analysis. >>>there must be something wrong with my F-test setting, or i must >>>misunderstanding the meaning of F-test in FSL... >>>Any help would be so appreciated!!! >>> >>>thanks