Mark,
thank you this clears up good part of my doubts (esp on the meaning of
the zstat).
I've been using zstats to compare across runs, but now I'm not sure how
this is biasing my comparisons, I'll look into doing my multivoxel
comparisons with the (CO)PEs.
thanks
martin
Mark Woolrich wrote:
> Martin,
>
>> 1) how is zstat different from tstat then (given they both are
>> defined as cope/sqrt(varcope))?
>
> Fair question. Strictly speaking tstat=cope/sqrt(varcope)). zstat is
> then obtained from the tstat by doing a probability transform. That
> is, the zstat is computed as the one which gives a P-value under the
> standardized normal distribution equal to the P-value from the
> t-distibution for the tstat.
>
>> 2) am I correct in interpreting your mail that comparing various
>> zstats obtained with a 'standard' FEAT high level (with the option of
>> grand mean scaling) session means comparing EVs' regression
>> coefficients (scaled such that m=0 and sd=1) [hence allowing me to
>> compare across runs/sessions]?
>
> Not quite. After doing grand mean scaling, the EV's regression
> coefficients (PEs or COPEs) themselves, without normalising them, can
> be used to compare across runs/sessions. This assumes that percent
> signal changes are the common currency to be compared between
> runs/sessions, and the use of grand mean scaling means that the PEs
> and COPEs are approximately proportional to percent signal changes
> with the same constant of proportionality between sessions/runs.
>
> Cheers, Mark.
>
> ----
> Dr Mark Woolrich
> EPSRC Advanced Research Fellow University Research Lecturer
>
> Oxford University Centre for Functional MRI of the Brain (FMRIB),
> John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
>
> Tel: (+44)1865-222782 Homepage: http://www.fmrib.ox.ac.uk/~woolrich
> <http://www.fmrib.ox.ac.uk/%7Ewoolrich>
>
>
>
>
> On 1 Nov 2006, at 19:06, Martin M. Monti wrote:
>
>> Mark,
>>
>> thank you. Two (final) questions:
>> 1) how is zstat different from tstat then (given they both are
>> defined as cope/sqrt(varcope))?
>> 2) am I correct in interpreting your mail that comparing various
>> zstats obtained with a 'standard' FEAT high level (with the option of
>> grand mean scaling) session means comparing EVs' regression
>> coefficients (scaled such that m=0 and sd=1) [hence allowing me to
>> compare across runs/sessions]?
>>
>> thank you for the clarification,
>>
>> martin
>>
>> Mark Woolrich wrote:
>>> Martin,
>>>
>>> zstat is COPE/sqrt(VARCOPE)
>>>
>>> To allow for higher-level analyses which compare EVs from different
>>> sessions (or runs), we strongly recommend that you do grand mean
>>> scaling, in which the whole 4D data set is normalised by a single
>>> scaling factor (each volume is scaled by the same amount), so that
>>> datasets from different sessions have the same grand mean. This
>>> controls for between session variability in the scaling of
>>> intensities the scanner spits out. The COPEs themselves should then
>>> be comparable between sessions.
>>>
>>> Grand mean scaling is the default in Feat.
>>>
>>> Cheers, Mark.
>>>
>>> ----
>>> Dr Mark Woolrich
>>> EPSRC Advanced Research Fellow University Research Lecturer
>>>
>>> Oxford University Centre for Functional MRI of the Brain (FMRIB),
>>> John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
>>>
>>> Tel: (+44)1865-222782 Homepage: http://www.fmrib.ox.ac.uk/~woolrich
>>> <http://www.fmrib.ox.ac.uk/%7Ewoolrich>
>>>
>>>
>>>
>>>
>>> On 1 Nov 2006, at 16:14, Martin M. Monti wrote:
>>>
>>>> Steve,
>>>>
>>>> I just now realized that when saying PE vs COPE I in fact meant
>>>> (CO)PE vs ZSTAT.
>>>>
>>>> Just to make sure I understand, zstat is COPE-mean(COPE)/sqrt(VARCOPE)?
>>>>
>>>> my point is that I compare patterns of activations across different
>>>> EVs, and some are collected in separate runs (with a run being
>>>> machine ON to machine OFF), hence comparing measures that do not
>>>> account for differences in signal mean and variability across runs
>>>> I may be getting non-interpretable results.
>>>>
>>>> cheers
>>>> martin
>>>>
>>>> Steve Smith wrote:
>>>>> Hi Martin,
>>>>>
>>>>> On 31 Oct 2006, at 20:35, Martin M. Monti wrote:
>>>>>
>>>>>> Dear Lords of FSL,
>>>>>>
>>>>>> I have a simple question re the contents of the PE and COPE images.
>>>>>>
>>>>>> At a first level (single sub - single run) PE1 is a matrix
>>>>>> containing the GLM parameter estimate (b) for each voxel for EV1.
>>>>>> If I created a contrast with (say) a 1 in EV1 and zeros in all
>>>>>> other EVs I would then get COPE1 to be in fact the same thing as
>>>>>> PE1 except it's a beta as opposed to a non standardized estimate
>>>>>> (i.e. b). [the betas are what I need..]
>>>>>>
>>>>>> So far so good (right? ..I think we previously discussed this
>>>>>> already..)
>>>>>
>>>>> Not quite - in this case (contrast [1 0 0...]) then COPE1 is
>>>>> _exactly_ PE1 - neither is standardised in any way at this point.
>>>>> PE1 is the beta, and for this contrast COPE1=PE1.
>>>>>
>>>>> I'm not sure what distinction you're making between b and beta. In
>>>>> the GLM:
>>>>> Y=Xb+e
>>>>> Then b (ie beta) is the parameter estimates.
>>>>>
>>>>>> Now, say that I have multiple runs and multiple subjects, and all
>>>>>> I do is average (i.e. no subtractions) EV1 across multiple
>>>>>> sessions (for each given subject) and across multiple subjects.
>>>>>> This group analysis will yield again a PE1 and a COPE1 (obtained
>>>>>> again just by creating a contrast with 1 in all EV1s and 0 for
>>>>>> all other EVs).
>>>>>>
>>>>>> Are the EP1 and COPE1 matrices just (group) average bs and
>>>>>> (group) average betas, respectively?
>>>>>
>>>>> Apart from the fact that the "bs" _are_ the betas, then yes, the
>>>>> rest is correct.
>>>>>
>>>>> Cheers, Steve.
>>>>>
>>>>>
>>>>>>
>>>>>> cheers
>>>>>> Martin
>>>>>>
>>>>>>
>>>>>> --
>>>>>>
>>>>>>
>>>>>>
>>>>>> --------------------------------------------
>>>>>> "Upward, not Northward!" [Edwin Abbott - FlatLand]
>>>>>> I became insane, with long intervals of horrible sanity. [E.A. Poe]
>>>>>> Would you tell me, please, which way I ought to go from here?
>>>>>> ''That depends a good deal on where you want to get to.'' Said
>>>>>> the Cat. I don't much care where -- Said Alice. Then it doesn't
>>>>>> matter which way you go, said the Cat. [Lewis Carroll]
>>>>>> Che si trattasse di una gara? ..Un colpo il mondo e un colpo tu,
>>>>>> a chi stupiva un po' di piu'.. [Per Un Vecchio Bambino - Roberto
>>>>>> Vecchioni]
>>>>>> --------------------------------------------
>>>>>> Martin M. Monti
>>>>>> Princeton University
>>>>>> Department of Psychology
>>>>>> Green Hall 3-S-8
>>>>>> (609) 258-5679
>>>>>> --------------------------------------------
>>>>>
>>>>>
>>>>> ---------------------------------------------------------------------------
>>>>> Stephen M. Smith, Professor of Biomedical Engineering
>>>>> Associate Director, Oxford University FMRIB Centre
>>>>>
>>>>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>>>>> +44 (0) 1865 222726 (fax 222717)
>>>>> [log in to unmask] <mailto:[log in to unmask]>
>>>>> http://www.fmrib.ox.ac.uk/~steve <http://www.fmrib.ox.ac.uk/%7Esteve>
>>>>> ---------------------------------------------------------------------------
>>>>
>>>> --
>>>>
>>>>
>>>>
>>>>
>>>> --------------------------------------------
>>>> "Upward, not Northward!" [Edwin Abbott - FlatLand]
>>>> I became insane, with long intervals of horrible sanity. [E.A. Poe]
>>>> Would you tell me, please, which way I ought to go from here?
>>>> ''That depends a good deal on where you want to get to.'' Said the
>>>> Cat. I don't much care where -- Said Alice. Then it doesn't matter
>>>> which way you go, said the Cat. [Lewis Carroll]
>>>> Che si trattasse di una gara? ..Un colpo il mondo e un colpo tu, a
>>>> chi stupiva un po' di piu'.. [Per Un Vecchio Bambino - Roberto
>>>> Vecchioni]
>>>> --------------------------------------------
>>>> Martin M. Monti
>>>> Princeton University
>>>> Department of Psychology
>>>> Green Hall 3-S-8
>>>> (609) 258-5679
>>>> www.webmartin.net
>>>> --------------------------------------------
>>>
>>
>> --
>>
>>
>>
>>
>> --------------------------------------------
>> "Upward, not Northward!" [Edwin Abbott - FlatLand]
>> I became insane, with long intervals of horrible sanity. [E.A. Poe]
>> Would you tell me, please, which way I ought to go from here? ''That
>> depends a good deal on where you want to get to.'' Said the Cat. I
>> don't much care where -- Said Alice. Then it doesn't matter which way
>> you go, said the Cat. [Lewis Carroll]
>> Che si trattasse di una gara? ..Un colpo il mondo e un colpo tu, a
>> chi stupiva un po' di piu'.. [Per Un Vecchio Bambino - Roberto Vecchioni]
>> --------------------------------------------
>> Martin M. Monti
>> Princeton University
>> Department of Psychology
>> Green Hall 3-S-8
>> (609) 258-5679
>> www.webmartin.net
>> --------------------------------------------
>
--
--------------------------------------------
"Upward, not Northward!" [Edwin Abbott - FlatLand]
I became insane, with long intervals of horrible sanity. [E.A. Poe]
Would you tell me, please, which way I ought to go from here? ''That depends a good deal on where you want to get to.'' Said the Cat. I don't much care where -- Said Alice. Then it doesn't matter which way you go, said the Cat. [Lewis Carroll]
Che si trattasse di una gara? ..Un colpo il mondo e un colpo tu, a chi stupiva un po' di piu'.. [Per Un Vecchio Bambino - Roberto Vecchioni]
--------------------------------------------
Martin M. Monti
Princeton University
Department of Psychology
Green Hall 3-S-8
(609) 258-5679
www.webmartin.net
--------------------------------------------
|