Dear Donald,
First of all thanks a lot for your quick reply. I have just realised
that making a contrast averaging the effect of all conditions across
sessions is quite simple, given that you enter all sessions in the same
model at 1st level (I use to consider them separatly hence my first
question).
In my exemple, I have 4 conditions (1 per session) and the informed
basis is used to model the HRF. Therefore at the first level the
parameters of my design are :
cond1_can cond1_tempodev cond1_dispdev cond2_can cond2_tempodev
cond2_dispdev cond3_can cond3_tempodev cond3_dispdev cond4_can
cond4_tempodev cond4_dispdev
where "can" stands for "canonical", "tempodev" for "temporal derivative"
and "dispdev" for "dispersion derivative"
Therefore for each subject I will have the following contrast to get the
average effect of all conditions : [1 0 0 1 0 0 1 0 0 1 0 0]
Thanks for the reference regarding my second question. I was aware of
this method as I read this paper before (thanks to this mailing list).
However, we decided to consider the derivatives separately at the second
level and to contrast the canonical to get the condition effects. I was
unsure on how to proceed to get the temporal and dispersion derivative
for this "average effect of all sessions" contrast. Now, I think that
the simplest solution is just to contrast the derivatives of all
conditions. According to my example, the contrast (for each subject)
would be : [0 1 0 0 1 0 0 1 0 0 1 0] for the temporal derivative and [0
0 1 0 0 1 0 0 1 0 0 1] for the dispersion derivative.
I hope that this make sense, please do not hesitate to correct me if
this is not valid,
Regards,
Camille
--
Camille MAUMET
Unité/Projet VisAGeS U746
IRISA-INRIA Rennes
Campus de Beaulieu
35042 Rennes Cedex
France
[log in to unmask]
MCLAREN, Donald wrote:
> See inline responses below.
>
> Best Regards, Donald McLaren
> =================
> D.G. McLaren
> University of Wisconsin - Madison
> Neuroscience Training Program
> Office: (608) 520-0586
> =====================
> This e-mail contains CONFIDENTIAL INFORMATION which may contain
> PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED
> and which is intended only for the use of the individual or entity
> named above. If the reader of the e-mail is not the intended recipient
> or the employee or agent responsible for delivering it to the intended
> recipient, you are hereby notified that you are in possession of
> confidential and privileged information. Any unauthorized use,
> disclosure, copying or the taking of any action in reliance on the
> contents of this information is strictly prohibited and may be
> unlawful. If you have received this e-mail unintentionally, please
> immediately notify the sender via telephone at (608) 520-0586 or email.
>
>
> On Fri, Oct 23, 2009 at 11:32 AM, Camille Maumet
> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>
> Dear SPMers,
>
> I really appreciated to find out information on how to compute
> group main
> effects in the context of a mixed design reading the thread entitled
> "flexible factorial - main effect of subject factor" (April 2009) (for
> instance:
> https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0904&L=SPM&P=R35471
> <https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0904&L=SPM&P=R35471>)
>
> This quote from Eric Zarahn outlines, in my point of view, one of
> the main
> points that came out from the discussion :
> | If you want to assess the classic main effect of "group" in the
> | context of a repeated measures ANOVA with "group" as a
> between-subject
> | factor and "condition" as a within-subject factor, then yes, in you
> | should create a contrast that averages over all conditions at
> the 1st
> | level, and then bring this to the second level.
>
> Basically I have two questions regarding the assessment of group main
> effects in this context :
> 1) Given that my conditions were acquired in 4 different sessions,
> I can't
> create a contrast that average across conditions at the first
> level using
> SPM contrast manager. In this case, is it valid to get the
> "average over all
> sessions" for each subject by simply averaging the con images
> (using imcalc
> for instance) obtained by contrasting each condition separately ?
>
>
> You can create contrasts across sessions -- assuming that your model
> has all four sessions in it. If you use 1 in each session, it will be
> 4 times the average of the confiles; but if you use .25 then you get
> the true average which is 1/4 of each session effect. Either way will
> work though. If you are contrasting 2 conditions -- then 1 -1 or .25
> -.25 would work. Sounds like you want the average of all conditions
> for all sessions, so you will want .125 for each condition in each
> session.
>
>
>
>
>
> 2) In addition, I modeled the haemodynamic response with the
> Informed Basis
> Set (canonical + temporal derivative + dispersion derivative). I
> wonder if
> it is valid to average the derivatives across sessions
> (independently for
> each subject) and bring them to the second level to gain
> statistical power ?
>
>
> See separate email.
>
>
> I was unsure whether I should continue the thread or start a new
> discussion... Hope that I was clear enough and would be more than
> happy to
> provide more details on my design if needed.
>
> Thanks in advance,
>
> Camille
>
>
|