[pymvpa] question about detrend and zscore

[John Clithero]

Hi, Matthias,

On Wed, Oct 7, 2009 at 8:44 AM, Matthias Ekman <Matthias.Ekman at nf.mpg.de> wrote:
> Hi,
>
> mh - may be I am wrong here, but from my point of view, high-pass
> filtering AND detrending is not "necessary". What happens if you
> zscore/detrend your data without step (4). My guess would be, that this
> somehow strange effect of negative correlation will disappear.

I think I may have written the wrong line of code earlier....if I just
use 'detrend' with 'model=constant', then that should not be a
problem, correct? And when I do that, then I still get this same
strange effect.


>
> Bests,
> Matthias
>
> John Clithero schrieb:
>> Hi,
>>
>> Looks like my email with the voxel image is waiting for moderator
>> approval, so here is my updated response without the picture:
>>
>> On Tue, Oct 6, 2009 at 4:39 PM, Yaroslav Halchenko
>> <debian at onerussian.com> wrote:
>>> On Tue, 06 Oct 2009, John Clithero wrote:
>>>> The 'wb_file' is func data with the relevant timepoints... it has had
>>>> some preprocessing done in FSL (motion correction, slice timing, brain
>>>> extraction). The ROI mask is a whole brain mask. It seems (I hope)
>>>> that the NiftiDataset is being put together correctly.
>>> wild guess - may be while doing FSL preprocessing you've done some
>>> intensity normalization?
>> Here (sorry if I didn't list them) are the preprocessing options I used in FSL
>> (1) brain extraction
>> (2) motion correction
>> (3) slice-timing correction
>> (4) high-pass filtering
>> Added this just now:
>> As far as I know, FSL does the following (from website and their
>> course lectures):
>> "Scale each 4D dataset by a single value to get the overall 4D mean
>> (dotted line) to be the same"
>> The intensitiy normalization was off during preprocessing (as FSL recommends).
>> Is this what you were asking about?
>>
>>>
>>>> This perfect negative correlation occurs even if I feed in arbitrary
>>>> labels to the NiftiDataset, so there must be some sort of error in how
>>>> I'm using detrend and/or score?? I am guessing this is my erros since
>>>> the raw feature data looks fine to me.
>>>> This perfect negative correlation also occurs after just implementing
>>>> "zscore" or "detrend", although obviously the values are different.
>>> so, they aren't present in just loaded dataset but if you do zscore or
>>> detrend -- they come?
>>
>> If I break data in half (A trials vs B trials, or random) before using
>> zscore or detrend, I generally see strong positive correlation (some
>> voxels are just more active than others across the experiment...this
>> makes sense).  But, once I use zscore or detrend....then goes to
>> perfect negative correlation (when looking at meanA vs meanB).
>> Each voxel after zscore, as it was said early, will have a mean of
>> zero across all trials...this is true for all the voxels.
>>
>>> could you just plot 1 voxel (which later carries perfect correlation)
>>> before/after detrend?
>>>
>> This is waiting for moderator approval, I guess.
>> Thanks for the help.
>>
>> Cheers,
>> John
>>
>>
>>> --
>>>                                  .-.
>>> =------------------------------   /v\  ----------------------------=
>>> Keep in touch                    // \\     (yoh@|www.)onerussian.com
>>> Yaroslav Halchenko              /(   )\               ICQ#: 60653192
>>>                   Linux User    ^^-^^    [175555]
>>>
>>>
>>>
>>> _______________________________________________
>>> Pkg-ExpPsy-PyMVPA mailing list
>>> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
>>> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
>>>
>>
>> _______________________________________________
>> Pkg-ExpPsy-PyMVPA mailing list
>> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
>> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
>