Hi Doug, did you already have a chance to look into this? Thanks, Caspar


2013/10/4 Caspar M. Schwiedrzik <cschwiedrz@rockefeller.edu>
Done. Thank you very much for looking into this.
Caspar


2013/10/4 Douglas N Greve <greve@nmr.mgh.harvard.edu>
Can you tar up you glmfit dir and drop it to me on our file drop?


On 10/04/2013 05:29 PM, Caspar M. Schwiedrzik wrote:
Hi Doug,

thank you very much for sending the Matlab function. When I run this, it creates a pcc.mgh file for my osgm contrast. However, the values seem strange. They range from -630 to 36 for my particular dataset.
I was expecting something between -1 and 1.
Caspar


2013/10/4 Douglas N Greve <greve@nmr.mgh.harvard.edu <mailto:greve@nmr.mgh.harvard.edu>>


    I think you are conflating the 1st level and the 2nd level. You
    could get pcc out of the 2nd level regardless of what you are
    using for the input from the first level. I've attached a matlab
    script that will compute the pcc for mri_glmfit output

    doug



    On 10/04/2013 03:33 PM, Caspar M. Schwiedrzik wrote:

        Hi Doug,
        I guess it boils down to the question how to get a group PCC
        map after a RFX GLM?
        Using -m PCC seems to only give me a map per subject. Are you
        calculating PCC from the t- values? Thanks,
        Caspar


        On Thursday, October 3, 2013, Caspar M. Schwiedrzik wrote:

            Hi Doug,

            On Thursday, October 3, 2013, Douglas N Greve wrote:


                It sounds like two issues:
                1. p-values not consistent with your program. What did
        you use
                to compute? Did you do a two-sided (which is what
        fsfast uses)?

            I used ttest in Matlab, two sided.

                2. Using pcc maps. Why not use -m pcc?


            Isn't that giving me a map per subject? How do I get the
        group map
            that is consistent with the results of mri_glmfit run on
        ces.nii?

            Thanks, Caspar


                doug


                On 10/03/2013 01:10 PM, Caspar M. Schwiedrzik wrote:

                    Hi Doug,
                    I loaded the pcc.nii file that I got from
        isxconcat-sess
                    into Matlab and then ran a t-test against 0 over
        the 4th
                    dimension. I converted the resulting p-values to
        -log10
                    and then compared them to the output of
        mri_glmfit, namely
                    sig.vol.
                    This was the mri_glmfit command:
                    mri_glmfit \
                    --surf averagesubject hemisphere \
                    --y pcc.nii \
                    --no-cortex \
                    --osgm \
                    --glmdir analysisname
                    I was expecting the p-values to be the same, which
                    apparently is not the case, unless I am
                    doing/understanding something wrong.

                    By now, I am actually more inclined to use the
        regression
                    coefficients instead. However, I'd still like to
        get pcc
                    maps from them, if there is a way to do so in FSFAST.
                    Thanks, Caspar




                    2013/10/3 Douglas N Greve
        <greve@nmr.mgh.harvard.edu <mailto:greve@nmr.mgh.harvard.edu>
                    <mailto:greve@nmr.mgh.harvard.edu
        <mailto:greve@nmr.mgh.harvard.edu>>>


                        On 10/03/2013 10:39 AM, Caspar M. Schwiedrzik
        wrote:

                            Hi Doug,

                            when I run a two-tailed t-test against 0
        in Matlab
                    on the Rs
                            in pcc.nii that I get out of
        isxconcat-sess with
                    -m pcc, and
                            DOF from ffxdof.dat, I get different -log10(p)
                    values than the
                            ones that come out of mri_glmfit.

                        I don't understand what you mean. Can  you
        elaborate?

                            I am not sure why this is happening.
                            In principle, I just want pcc maps as
        final output
                    to show
                            them on the surface (instead of p-values).
        So I'd
                    be happy to
                            follow your advice regarding the biasing
        effects
                    of noise and
                            autocorrelation and use the regression
                    coefficients. However,
                            mri_glmfit (v5.1) does not seem to output
        pcc maps
                    of the
                            contrasts (contrary to selxavg3-sess on
        the single
                    subject
                            level). How would I get those?

                            Thanks, Caspar


                            2013/10/1 Douglas N Greve
        <greve@nmr.mgh.harvard.edu <mailto:greve@nmr.mgh.harvard.edu>
                            <mailto:greve@nmr.mgh.harvard.edu
        <mailto:greve@nmr.mgh.harvard.edu>>
                            <mailto:greve@nmr.mgh.harvard.edu
        <mailto:greve@nmr.mgh.harvard.edu>
                            <mailto:greve@nmr.mgh.harvard.edu
        <mailto:greve@nmr.mgh.harvard.edu>>>>



                                On 10/01/2013 01:13 PM, Caspar M.
        Schwiedrzik
                    wrote:
                                > Hi Doug,
                                > it would be great if you could give
        me some
                    further
                            advise on the
                                > group analysis of functional
        connectivity maps.
                                > Specifically, I am trying to get PCC
        maps
                    for certain
                            seeds, and am
                                > not planning any comparison between
        groups.
                                > Following your previous advise, I am
        running
                            isxconcat-sess with -m
                                > pcc to get the PCC maps.
                                > I would then run
                                >
                                > mri_glmfit \
                                > --surf averagesubject hemisphere \
                                > --y pcc.nii \
                                > --no-cortex \
                                > --osgm \
                                > --glmdir analysisname
                                >
                                > *Could you please provide some more
        detail
                    on what kind of
                                analysis is
                                > performed when I provide pcc.nii as
        an input for
                            mri_glmfit? Is it a
                                > t-test of the Fisher-transformed
        r-values
                    against 0?
                                I just run a t-test of the r-values. I
        don't
                    have a
                            program to convert
                                them to z-values, however, there are
        z-values
                    that are
                            created in the
                                first level analysis. These are
        generated from the
                            p-values but I
                                bet it
                                would give you the same thing. Use -m
        z with
                            isxconcat-sess if you
                                want
                                to use the z.
                                > *Is the average r-value or z-value saved
                    somewhere?
                                Which level? For mri_glmfit,  they are
        not,
                    but it is not
                            hard to get
                                them with matlab.
                                > *Do you take the autocorrelation into
                    account (as in
                            Vincent JL et
                                > al., 2007. Intrinsic functional
        architecture
                    in the
                            anaesthetized
                                > monkey brain. Nature. 447:83-86)?
                                Not usually, but it could be done by
        not including
                            -no-whiten when you
                                run mkanalysis-sess. I usually use the
        regression
                            coefficients instead
                                of correlation coefficients because
        that they
                    are at least
                                unbiased with
                                respect to noise level and
        autocorrelation.
                                doug


                                > I'd also be happy to look this up
        but I'd
                    need to know
                            where I can


    --     Douglas N. Greve, Ph.D.
    MGH-NMR Center
    Outgoing:
    ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/



    The information in this e-mail is intended only for the person to
    whom it is
    addressed. If you believe this e-mail was sent to you in error and
    the e-mail
    contains patient information, please contact the Partners
    Compliance HelpLine at
    http://www.partners.org/complianceline . If the e-mail was sent to
    you in error
    but does not contain patient information, please contact the
    sender and properly
    dispose of the e-mail.