Hi Simon
> Hi Cyril,
> Thanks very much for the prompt reply.
> > 1st thing is to change your ANOVA, I think you have a problem there.
> > Using SPM5/8 choose flexible factorial and enter 'subject' for your 1st
> > factor, this name is a special name allowing SPM to know you are
> doing a
> > repeated measure, then enter condition for your 2nd factor, and finally
> > fill 2 cells with the corresponding images - this should allow you to
> > build a proper model giving you almost identical results as your paired
> > t-test
> Yes, I did a flexible factorial model as well, and certainly that's
> what I would use in practice if I wanted a repeated measures anova at
> the second level, but in that case the extra sensitivity through
> proper modelling of subject variations within the design matrix gives
> a different (generally more active) t-map, hence it can't be entered
> into a direct comparison, which is why I excluded it. I chose to
> model the subjects with covariates instead in order to be exactly
> equivalent to the paired t-test and one-sample t-test approaches, and
> the fact that the t-maps were identical suggests this is the case?
ok I get the picture now .. are your t-maps that identical? you can
formally test this using ImCalc and compute the difference between two t
maps (i1-i2) looking at this map of difference will tell you where it
differs ..
>
> > finally one sample on the difference vs paired/ANOVA on each condition
> > --> well results should not be that different .. but beyond this it is
> > your own choice - this is a theoretical choice - the main difference
> > being that in the later case you model the covariance between
> conditions
> > and weight the solution whereas in the 1st case you don't.
>
> Okay, that makes sense, but I'm not sure it answers my fundamental
> question: if the three t-maps are identical, why aren't the three sets
> of eigenstatistics taken from the same cluster in the three t-maps
> also identical? Or to put it another way, what extra information goes
> into the cluster eigenstatistics that doesn't go into the t-maps, and
> where does it come from?
it they were identical the eigen values would also be - I can see two
explanations, 1) the cluster you extract the data from is not the same
and 2) some values in some maps are extreme ... you could check all this
using a small tool I wrote (Easy_ROI - http://www.sbirc.ed.ac.uk/cyril)
by taking the cluster from e.g. the one sample map and plot / extract
the eigen values from the one sample, paired and ANOVA maps - note that
the file saved by Easy_ROI contains the raw values, the mean and the
eigen values so you could really compare it all
hope this helps
cyril
>
>
>
> > Date: Mon, 16 Aug 2010 10:41:20 +0100
> > From: [log in to unmask]
> > Subject: Re: [SPM] One-sample t-test, paired_sample t-test, one-way
> anova: same t-maps, different eigenvariate stats?
> > To: [log in to unmask]
> >
> > Dear Simon
> >
> > 1st thing is to change your ANOVA, I think you have a problem there.
> > Using SPM5/8 choose flexible factorial and enter 'subject' for your 1st
> > factor, this name is a special name allowing SPM to know you are
> doing a
> > repeated measure, then enter condition for your 2nd factor, and finally
> > fill 2 cells with the corresponding images - this should allow you to
> > build a proper model giving you almost identical results as your paired
> > t-test
> >
> > 2nd you mentioned covariates - not that in SPM5/8 you can enter
> > covariates with your paired t-test as well ..
> >
> > finally one sample on the difference vs paired/ANOVA on each condition
> > --> well results should not be that different .. but beyond this it is
> > your own choice - this is a theoretical choice - the main difference
> > being that in the later case you model the covariance between
> conditions
> > and weight the solution whereas in the 1st case you don't.
> >
> > Hope this helps
> > Best
> > Cyril
> >
> >
> > > Dear SPMers,
> > >
> > > I'm hoping someone can educate me a bit on something I've noticed
> when
> > > playing around with a recent analysis; I'm sure there's a good reason
> > > for it but my understanding seems to be a bit hazy in this area....
> > >
> > > I have a group of 36 subjects, which for the sake of the current
> > > question we will regard as a single group. In the first level design
> > > matrix for each subject we have condition 1, condition 2 (two levels
> > > of a single factor) and condition 3 (button press no-interest
> > > regressor), plus movement regressors and a baseline.
> > >
> > > I'm comparing three different ways of analysing them in SPM8:-
> > >
> > > (1) Method 1 - One-sample t-test: first-level contrast 'Condition 1 >
> > > Condition 2', which gives me one image per subject taken forward to
> > > the second level. This is entered into one-sample t-test, which gives
> > > me a t-map.
> > >
> > > (2) Method 2 - Paired-sample t-test: first-level contrasts 'Condition
> > > 1' and 'Condition 2' separately for each subject, which gives me two
> > > images per subject taken forward to the second level. These are
> > > entered into a paired-samples t-test, which gives me a t-map.
> > >
> > > (3) Method 3 - One-way within-subjects anova: first-level contrasts
> > > 'Condition 1' and 'Condition 2' separately for each subject, which
> > > gives me two images per subject taken forward to the second level.
> > > These are entered into a one-way within-subjects ANOVA, which also
> > > includes subject regressors/covariates in the way recommended in the
> > > Henson and Penny paper (and the equivalent chapter in the SPM book).
> > > Now as it happens, there is no activation for the case of
> 'Condition 2
> > > > Condition 1', so my F-map for the Main Effect of Factor 1, and the
> > > t-map for Positive Effect of Factor 1, are identical.
> > >
> > > As far as I can see, and premised on there being no deactivation in
> > > Method 3, these three methods are essentially equivalent, although
> > > I've attached the design matrices for all three in case I've made a
> > > mistake that someone can tell me about. Sure enough, the t-maps
> > > obtained from all three methods are absolutely identical (I've done a
> > > voxel-wise comparison of all three to confirm this).
> > >
> > > So far so good; now here comes the fun part. If I extract the first
> > > eigenvariate statistics for a given cluster, these are completely
> > > different for all three methods, in spite of the t-images being
> > > identical. Obviously the one-sample t-test at least needs to be
> > > different because it has just 36 values fo xY.u (because of 36
> images)
> > > while the paired t-test and the anova methods have 72, but I can find
> > > no obvious relationship between any of them. I've attached maps of
> > > xY.y from the three different methods for the same cluster for
> > > information purposes.
> > >
> > > So, what is the relationship between the cluster statistics obtained
> > > from these three seemingly equivalent methods? Which should I be
> > > using? I realise that Method 1 is the most common approach, but I'll
> > > be adding some further (between and within subjects) factors into the
> > > design shortly, so I'd prefer an anova-based approach if possible.
> > >
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|