Hello Burkhard,
In your example the raw data consists of FA values, so the only question that the one-sample test can answer is if the group's FA score is >0 on average. If you subtracted the global mean FA from your input data, then the one sample test could be used to find where the FA-difference was significant. If your higher-level input consisted of (e.g.) t-values with a known distribution, then the mean group effect is convertible to p-values ( without using a non-parametric method like randomise ).
Many Regards
Matthew
> Hi Matthew,
> sorry but I am confused ( I am not a stats person and the lingo sometimes is a bit overwhelming).
>
> The help on "Randomize v2.1" clearly states the one-sample t-test: One sample t-test on difference measures!
>
> Furthermore in the FEAT-manual (nearly at the end before the design matrices are displayed):
> Singe group average (One Sample T-test):
> "We have 8 subjects all in one group and want the mean group effect. DOES THE GROUP ACTIVATE ON AVERAGE?"
>
> If we just test if the mean is different from Zero we won't get an information on "activation". In a BOLD
> experiment the signal mean is always >0 (basically for any normal MR-data the mean is always >0 because we are
> working with modulus images - even regions of "pure" noise have a mean >0!.
> So how does this test evaluate activation pattern in a single group?
>
> sorry if I appear stubborn but I would really like to understand the meaning for the randomize one-sample
> t-test!
>
> Cheers,
> Boogie
>
|