Hi Anderson,
you have been very clear.
Can I ask some other 2 questions (sorry).
I understand everything, however:
1) Can I run the Repeated measures ANOVA to have the f-stat, and then the "Tripled" T-Test to have the t-stats?
2) I can use PALM, to have the multiple testing corrections, however, if I use RANDOMISE and then I correct the t-stat, for instance by Bonferroni, is it a similar thing?
Thanks again,
Maurizio
> Hi Maurizio,
>
> The tripled t-test is in fact a kind of ANOVA, except that you do the pairwise comparisons, as opposed to an F-test.
>
> For the last question: yes and no. It's the same situation we face when we have, say, 3 groups, A, B and C. For a comparison between A and B only, should C remain in the model? It's a matter of opinion. I would say yes, and here are the reasons:
> - All comparisons will be made, and in it, assumptions such as that errors have homogeneous variances will be done for each pairwise test, so you may as well make that globally and put all participants together.
> - The estimate of that variance will be better because it will be based on a larger sample size.
> - Correction for multiple testing becomes straightforward using permutations.
> However, specifically in the case of repeated measures, the paired solution has one benefit that makes it attractive: compound symmetry always holds, whereas for the 3 measures as in the tripled t-test it may fail. To correct in this case, you can use PALM. That is, you'd do 3 paired t-tests, and use -corrmod and -corrcon to correct across. The F-test (that can be done with the triple t-test) can be replaced, with advantages, for NPC. If you decide to take this route and need further details on how to set it up, please feel free to post again.
> All the best,
> Anderson
> On Sun, 8 Sep 2019 at 09:59, Maurizio Bergamino <[log in to unmask]> wrote:
> Thanks Anderson,
> therefore, your suggestion is to use the Single-Group- Three Measurements ("Tripled T-Test") and not ANOVA?
> And... sorry last question, is there any difference if I use the tripled-T-test or different paired t-tests?
> Maurizio
> Hi Maurizio,
> The example that you really need is this: https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#Single-Group.2C_Three_Measurements_.28.22Tripled_T-> Test.22.29.
> > You can in fact run this model in randomise: define one exchangeabiliy block per participant, and supply it with the option -e. That should be it.
> All the best,
> Anderson
>> On Tue, 3 Sep 2019 at 10:06, Maurizio B. <[log in to unmask]> wrote:
>> Hello FSL experts,
>> I would like to analyze DTI maps in a group of patients (N=41) scanned in three different time-points (baseline, after 24 months, after 48 months).
>> I would like to know if I have differences across the three time-points (Baseline vs TP1) (Baseline vs TP2), and (TP2 vs TP3).
>> I think that I have to use Repeated measures ANOVA but I am not sure.
>> I am also not sure about the matrices that I have to use for Randomise.
>> I have attached the matrices that I would like to use.
>> Please, can someone helps me to understand if repeated measures ANOVA and my matrices are fine?
>> Second question:
>> In the webpage of Repeated measures ANOVA (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Randomise/UserGuide#Repeated_measures_ANOVA) There >> is an example with 1-factor 4-levels (Repeated Measures). My case should be 1-factor 3-levels (Repeated Measures).
>> I do not understand why the three t-contrasts for the categorical effect are
>> 0 0 1 0 0
>> 0 0 0 1 0
>> 0 0 0 0 1
>> Therefore only for the first 3 levels (and not for all 4!).
>> Thanks,
>> Maurizio
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
|