Hi,

when you set up the relevant contrasts for a triple t-test at the  
highest level this will automagically be calculated for every lower- 
level contrast. That is, at the highest level you do not need to  
specify any of the event types - after feeding in all the lower-level  
feat directories in the higher-level GUI a set of tickboxes should  
appear which allow you to select the lower-level contrasts you want  
to have the higher-level comparison calculated on.

The higher-level analysis is calculated using the unthresholded lower- 
level results.
hope this helps
Christian




On 20 Jul 2007, at 19:16, Steward Carolyn wrote:

> Hi,
>
> Thanks very much for your swift reply! I've had a go at inputting feat
> directories of individual subjects where both event types have been
> included as separate EVs but can't see how the event types are
> distinguished between when you put them into higher level analysis. Is
> there a particular way of setting up the EVs at the higher level so  
> that
> the event types can be differentiated?
> Also, do the post-stats thresholds set at the individual subject level
> affect stats done at higher levels?
>
> Many thanks,
> Carolyn
>
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf Of Christian Beckmann
> Sent: 20 July 2007 16:33
> To: [log in to unmask]
> Subject: Re: [FSL] event-related fMRI analysis
>
> Hi
>
>> In my study there were 16 subjects who were scanned on 2 separate
>> days. On day 1 they performed a cognitive task at baseline and again
>> following the administration of a placebo or drug, and on day 2 this
>> was repeated, such that each subject performed the task 4 times: at
>> baseline ('BP') then placebo ('P') and at baseline ('BD') then drug
>> ('D'). The task consisted of 2 event types A and B, and reaction  
>> times
>
>> (RTs) were obtained for each event.
>>
>> Firstly, I would like to use RT as a covariate. My questions are the
>> following:
>> 1. Can I combine the 2 different event types with their corresponding
>> RTs in the same analysis or do they have to be analysed separately?
>> i.e. will
>> I need to analyse event type A for each subject and feed these feat
>> directories into higher level analysis and repeat this for event type
>> B?
>>
>
> No, if your experiment involved two different event types you should
> model both of these in a single lower-level design. Each event type  
> will
> become a separate EV. RT information will also be a separate EV.
>
>> 2. Will I need to analyse the sessions (BP, P, BD and D)  
>> separately or
>
>> can each individual and each session be put into the same higher  
>> level
>
>> analysis with RT as a covariate?
>
> I guess the simplest would be to model this as a triple t-test, see
> http://www.fmrib.ox.ac.uk/fsl/feat5/index.html
>   At an intermediate level you'd then use a fixed-effects model to
> combine the two BP sessions.
>
>> 3. I'm unclear as to what people mean by demeaning covariates, please
>> can you explain?!
>
> You remove the mean value from the time series of RTs -  in the GLM as
> applied to fMRI you use covariates to 'explain' intensity changes  
> in the
> data.The mean image intensity is of no interest (at the first
> level) and therefore is removed, i.e. the data is mean zero.
> Therefore, all the covariates should be zero mean, too.
>
>> 4. I'm also not sure whether it would be appropriate to tick the
>> orthogonalise boxes?
>>
>
> I suggest you do not orthogonalise the RT regressor wrt any of the  
> other
> ones - if you do and your EVs are positively correlated with RTs  
> you end
> up boosting the EVs for A and/or B because by orthogonalising RTs any
> amount of variance which could be explained either by A/B or RT  
> ends up
> being attributed to A/B only.
>
>> I'd also like to look at the mean group BOLD responses in the 4
>> different conditions, and compare between them (e.g. D versus BD, D
>> versus P).
>> 1. Again, can I combine the 2 different event types in the same
>> analysis or do they have to be analysed separately?
>
> see above
>
>> 2. Which is the best approach to do this, to put all data (each
>> subject and each session) into a one factor 4 level repeated measures
>> ANOVA?
>>
> I suggest triple t-test, , just specify the appropriate contrasts to
> compare e.g. D to P hope this helps Christian
>
>> A couple of other things, do the post-stats set at the individual
>> subject level affect stats done at higher levels? Also, I'm confused
>> about what the difference is between inputting feat directories or
>> cope images into higher level analysis.
>>
>> Many thanks!
>> Carolyn
>
> ____
> Christian F. Beckmann
> University Research Lecturer
> Oxford University Centre for Functional MRI of the Brain (FMRIB) John
> Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
> [log in to unmask]	http://www.fmrib.ox.ac.uk/~beckmann
> tel: +44 1865 222551			fax: +44 1865 222717
>
> This message has been checked for viruses but the contents of an  
> attachment
> may still contain software viruses, which could damage your  
> computer system:
> you are advised to perform your own checks. Email communications  
> with the
> University of Nottingham may be monitored as permitted by UK  
> legislation.

____
Christian F. Beckmann
University Research Lecturer
Oxford University Centre for Functional MRI of the Brain (FMRIB)
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
[log in to unmask]	http://www.fmrib.ox.ac.uk/~beckmann
tel: +44 1865 222551			fax: +44 1865 222717