Subject: | | Re: question on spm5 |
From: | | [log in to unmask][log in to unmask]> Unidad de Trastornos de Memoria. Neurologia Clinica Universitaria de Navarra Pamplona (Spain)38_5Jul200710:55:[log in to unmask] |
Reply-To: | | [log in to unmask][log in to unmask] >>> Universitaetsklinikum Freiburg Phone 49(0)761-270-5331 >>> Breisacher Str. 64 Fax 49(0)761-270-5416 >>> 79106 Freiburg http://fbi.uniklinik-freiburg.de/ >>> ========================================================================= >>> >>> >> >> >> >> > ------------------------------------------------------------------------ > > function con = spm_config_contrasts > % Configuration file for contrast jobs > %_______________________________________________________________________ > % Copyright (C) 2005 Wellcome Department of Imaging Neuroscience > > % Darren Gitelman > % $Id: spm_config_contrasts.m 802 2007-04-27 07:47:35Z volkmar $ > > > %_______________________________________________________________________ > > > spm.type = 'files'; > spm.name = 'Select SPM.mat'; > spm.tag = 'spmmat'; > spm.num = [1 1]; > spm.filter = 'mat'; > spm.ufilter = '^SPM\.mat$'; > spm.help = {'Select SPM.mat file for contrasts'}; > > sessrep.type = 'menu'; > sessrep.name = 'Replicate over sessions'; > sessrep.tag = 'sessrep'; > sessrep.labels = {'Don''t replicate','Replicate','Create per session','Both'}; > sessrep.values = {'none','repl','sess','both'}; > sessrep.val = {'none'}; > sessrep.help = {['If there are multiple sessions with identical conditions, ' ... > 'one might want to specify contrasts which are identical over ',... > 'sessions. This can be done automatically based on the contrast '... > 'spec for one session.'],... > ['Contrasts can be either replicated (thus testing average ' ... > 'effects over sessions) or created per session. In both ' ... > 'cases, zero padding up to the length of each session ' ... > 'and the block effects is done automatically.']}; > > name.type = 'entry'; > name.name = 'Name'; > name.tag = 'name'; > name.strtype = 's'; > name.num = [1 1]; > name.help = {'Name of contrast'}; > > tconvec.type = 'entry'; > tconvec.name = 'T contrast vector'; > tconvec.tag = 'convec'; > tconvec.strtype = 'e'; > tconvec.num = [1 Inf]; > tconvec.help = {[... > 'Enter T contrast vector. This is done similarly to the ',... > 'SPM2 contrast manager. A 1 x n vector should be entered ',... > 'for T-contrasts.']}; > > fconvec.type = 'entry'; > fconvec.name = 'F contrast vector'; > fconvec.tag = 'convec'; > fconvec.strtype = 'e'; > fconvec.num = [Inf Inf]; > fconvec.help = {[... > 'Enter F contrast vector. This is done similarly to the ',... > 'SPM2 contrast manager. One or multiline contrasts ',... > 'may be entered.']}; > > fconvecs.type = 'repeat'; > fconvecs.name = 'Contrast vectors'; > fconvecs.tag = 'convecs'; > fconvecs.values = {fconvec}; > fconvecs.help = {... > 'F contrasts are defined by a series of vectors.'}; > > tcon.type = 'branch'; > tcon.name = 'T-contrast'; > tcon.tag = 'tcon'; > tcon.val = {name,tconvec,sessrep}; > tcon.help = {... > '* Simple one-dimensional contrasts for an SPM{T}','',[... > 'A simple contrast for an SPM{T} tests the null hypothesis c''B=0 ',... > 'against the one-sided alternative c''B>0, where c is a column vector. '],'',[... > ' Note that throughout SPM, the transpose of the contrast weights is ',... > 'used for display and input. That is, you''ll enter and visualise c''. ',... > 'For an SPM{T} this will be a row vector.'],'',[... > 'For example, if you have a design in which the first two columns of ',... > 'the design matrix correspond to the effects for "baseline"MøX¸ |
Date: | | Thu, 12 Jul 2007 08:20:34 +0100 |
Content-Type: | | text/plain |
Parts/Attachments: |
|
|
|
|
Hi Roberto
> Dear all,
> this must be a basic question for must of you, but I need your qualified
> opinion.
well let see
> I am running an spm analysis with a cannonical 2 by 3 design, and I wonder
> which of these two approaches is the appropriate:
> After specifying the matrix design with six conditions (i.e., A1, A2, A3, B1,
> B2, B3) I might run for the first level analysis:
> A1: 1 0 0 0 0 0
> A2 0 1 0 0 0 0
> . 0 0 1 0 0 0
> . 0 0 0 1 0 0
> . 0 0 0 0 1 0
> A6 0 0 0 0 0 1
> which represents the averages on each condition per subject
> at the second level, then, I might contrast, for example, A1>B1 with the *com
> images in a full factorial anova
> the second approach would be doing direct contrasts at the first level:
> E.g, A1>B1 1 0 0 -1 0 0 per subject, and then, performing
> per-specific-contrast one sample t tests at the second level with all the
> respective subject's com images.
> which of these two approaches is the most appropriate? I think the second one
> is more sensitive per specific comparisons while keeping the advantages of
> the random effects test. is not it?
both approaches are valid and correspond to e RFX analysis, now you may gain
some sensitivity with approach 2 but loose flexibility - to my opinion the best
is to set up the full factorial anova as you can then look at A (111) B
(000111) obvisously A vs B, at all simple effects A1 vs B1, A2 vs B2 and B3 vs
B3 but more importantly at the interactions between these factors eg A1>B1 but
A2<B2 .. (1-10-110) - I guess it really depends on what question you want to
ask - if your interest is always in A>B then yes compute for each subject
A1>B1, A2>B2 and A3>B3 then enter those 3 contrasts in an ANOVA ...
hope this helps
cyril
------------------------------------------------------------------
Message sent via Psychology Dept. WEB-Mail Gateway.
http://www.psy.gla.ac.uk/
------------------------------------------------------------------
|
|
|
|