Is spatial mm implemented in a user-friendly way in FSL yet?!
Cheers,
Cornelius
Am Montag, den 31.01.2005, 13:50 +0100 schrieb Andreas Bartsch:
> Hi,
> it does depend on the size and stability of the effect. Both may be rather pronounced in motor tasks but lower in complex cognitive tasks. In any case, 20 - 30 trials are not a lot. Twice as many would be much better. You may be propelled to use spatial mixture modelling, for example, and you should feel ready and comfortable to look at you data in this way (providing you direct control about the true positive rate). If you want / have to stick to conventional inference about ME (controlling the false positive rate), you are probably better off increasing your trial numbers.
> Cheers-
> Andreas
>
>
> ________________________________
>
> Von: FSL - FMRIB's Software Library im Auftrag von Hanne Lehn
> Gesendet: Mo 31.01.2005 13:20
> An: [log in to unmask]
> Betreff: [FSL] #=? trials for sufficient stat power
>
>
>
> Hi,
>
> I am in the process of designing an event-related experiment, and currently
> pondering about this basic question: How many trials do I really need?
>
> I have 2 experimental conditions á 50 trials (SD=2500 ms). Each condition is
> further categorized according to the subjects' responses (correct or
> incorrect). Assuming an accuracy rate of about 50-60%, the relevant response
> categories (which define my main contrast) will contain 20-30 trials each. I
> wonder if this will be sufficient to observe statistically significant
> effects... The magnitude of bold changes in the targeted area is fairly low
> and shows much variation, which argues for more repetitions of critical
> events. Unfortunately, increasing the number of trials is difficult in my
> case, as the total scan time is already at what I consider to be a maximum.
>
> I have not yet come across many proper investigations of the relationship
> between number of trials and statistical power in fmri analysis. So: Does
> any of you have ideas about the minimum number of trials required in a
> design like mine? Perhaps had a look at this in your own data?
>
> Thanks!
>
> Hanne
>
--
Cornelius Werner
Institut fuer Medizin (IME)
AG Kognitive Neurologie
Forschungszentrum Juelich
52425 Juelich
Germany
Tel. +49-(0)2461-61-8609
|