Rik,

        Thanks for looking into this for me.  I downloaded the attached file into
the SPM99 directory and tried to specify the model but the same error
reoccurred.  I double-checked to be certain that the version of the
spm_get_ons.m file you sent wasn't the same as the version that was
originally in the directory and it wasn't.  Interestingly, I tried to
specify the exact same model in SPM99 on a linux box we have here in the lab
and it worked fine.  Previously I had been working on a Win2000 box.  I then
tried copying the version of the spm_get_ons.m file that was on the linux
machine to the Win2000 box, but the problem persists.  So, for the moment I
can continue analyzing the dataset (thanks to our linux box), but the
mystery of why the model can't be specified on the Win2000 machine remains
unresolved.

Any other ideas?

- Kyle

-----Original Message-----
From: Rik Henson [mailto:[log in to unmask]]
Sent: Thursday, September 25, 2003 8:41 AM
To: Kyle Simmons
Cc: [log in to unmask]
Subject: Re: parametric modulation errors


Kyle -

I replicated your problem - I think there is a bug in SPM99's version
of spm_get_ons.m (that probably only arises with this type of parametric
modulation). Try the attached version instead - it worked fine for me
(using your first option below).

Rik




Kyle Simmons wrote:
>
> Dear SPMers,
>
>  I am attempting to specify a model using parametric modulation in
> SPM99.  Unfortunately, I've been running into several roadblocks.  First,
> let me describe the design, and then I'll describe the problems.
>  For each subject there are 5 runs, each run having 114 scans and a
> TR=2.5.  The conditions are replicated in each run, but the timing
> parameters vary by run as a function of the parametric modulation.  There
> are three conditions: rest, lexical decision, and property generation.  In
> all runs the onset vectors are as follows:
>      Rest = 1, 97
>      Lexical decision = 15, 43, 71
>      Property Generation = 29, 57, 85
> The parametric modulation pertains to the property generation condition.
> There are three types of words about which subjects should generate
> properties: low word associate words, medium word associate words, and
high
> word associate words.  The order in which the three word associate levels
> are presented is counterbalanced across runs.  For example, on some runs
> the low word associate condition appears at scan 29, on others it might
> appear at scan 85, and so on for each word associate level.
>
> This now brings me to my problems.
> 1)  When I enter Rest as the first condition, lexical decision as the
> second condition, and property generation as the third condition, SPM
> crashes on me as soon as it asks "which trials" the parametric modulation
> variable should pertain to.  The input of events looks something like the
> following:
>   - Vector of onsets for Rest = 1 97
>   - Variable durations = no
>   - Vector of onsets for LexicalDecision = 15 43 71
>   - Variable durations = no
>   - Vector of onsets for PropertyGeneration = 29 57 85
>   - Variable durations = no
>   - Parametric modulations = other
>   - Name of parameter =  WA_level
>   - Expansion =  linear
>   - Which Trials =  3
> At this point all of SPM freezes.  When I look in the Matlab window I see
> the following error messages:
>
>      SPM99: spm_fmri_design (v2.27)             11:46:41 - 23/09/2003
>      ----------------------------------------------------------------
>      ??? Index exceeds matrix dimensions.
>
>      Error in ==> C:\matlab_sv12\spm99\spm_get_ons.m
>      On line 399  ==>                 ns    = length(on{i});
>
>      Error in ==> C:\matlab_sv12\spm99\spm_fMRI_design.m
>      On line 215  ==>   [SF,Cname,Pv,Pname,DSstr] = ...
>
>      Error in ==> C:\matlab_sv12\spm99\spm_fmri_spm_ui.m
>      On line 206  ==>  [xX,Sess] = spm_fMRI_design;
>
>      ??? Error while evaluating uicontrol Callback.
>      ----------------------------------------------------------------
>
>      I've downloaded all the most recent patches and this doesn't seem to
> fix anything.  Any ideas about how to fix this?
>
> 2)  As a work-around for this problem I tried to enter all three
conditions
> into the design, but this time entering the property generation condition
> as the first condition, with lexical decision and rest being the second
and
> third conditions.  In this case, I am able to get past the previously
> described error, but a new error arises.  Now, when it asks for the
> parameters for the property generation variable (i.e., the onset vectors
> indexing the low, medium, and high word associate blocks), it will only
> allow me to enter 2 vectors, rather than the three that I need to enter.
> In this case, the input of events looks like the following:
>   - Vector of onsets for PropertyGeneration = 29 57 85
>   - Variable durations = no
>   - Vector of onsets for lexdec = 15 43 71
>   - Variable durations = no
>   - Vector of onsets for rest = 1 97
>   - Variable durations = no
>   - Parametric modulations = other
>   - Name of parameter =  WA_level
>   - Expansion =  linear
>   - Which Trials =  1
>   - [2] Parameters for PropertyGeneration =
> When I try to enter 57 29 85 (which shows that the low word associate came
> at scan 57, medium at 29, and high and 85) I get an error saying "2
> vector required".  Any suggestions why this might be the case?
>
> Thanks,
>  Kyle

--
---------------------------------------------------------

DR RICHARD HENSON
Institute of Cognitive Neuroscience
& Wellcome Department of Imaging Neuroscience
University College London
17 Queen Square
London, WC1N 3AR
England

EMAIL:  [log in to unmask]
URL:    http://www.fil.ion.ucl.ac.uk/~rhenson

TEL1    +44 (0)20 7679 1120
TEL2    +44 (0)20 7833 7472
FAX     +44 (0)20 7813 1420
MOB     +44 (0)794 1377 345

---------------------------------------------------------
--