On Wed, 26 Jan 2005 15:41:46 +0900, Archana Singh
<[log in to unmask]> wrote:
>Hi,
>This is for Karl or anyone who likes to respond.
>I am quoting the portion of an spm_manual on summary-stat appraoch for RFx
>as reference for my question.
>According to the explanation, this means that we summarize subject-wise
>means in a single sample and do a single
>sample t-test. It suggests that variance of the summarized means itself
>would reflect both within-subject
> and between-subject variance. Does it mean we don't need to know
>within-subject variance from subjects
>at the second level?
The within-subject variance doesn't appear *explicitly* in the summary-
statistic method. But the summary-statistics method does implicitly
include it.
See eqn (19) of
http://www.fil.ion.ucl.ac.uk/spm/doc/books/hbf2/pdfs/Ch12.pdf
HTH,
S
> I am not sure whether I understand it correctly.
>I am trying this approach with data obtained from functional NIRS(near
>infrared spectroscopy)
>and the results do not conform to this explanation.
>I do a t-test with two different denominators as follows:
>1) std err of summarized samples
>2) std err derived from both between-subject and within-subject variance
>(varbs/N + varws/Nn) as explained in RFx paper by Karl.
>And results are far from identical. I wonder why. In my data from NIRS,
>within-subject variance is considerably high and almost comparable to
>between-subject variance. Also, the size of the sample is small, i.e. 5.
>Thanks for your response,
>Archana
>______________________________From spm_RandFX.man___________________
>%The residual error variance of this second
>% level model is the variance of the contrast images from subject to
>% subject, and consists of contributions from both the between and
>% within subject components of variance, in the correct proportions. It
>% can easily be shown that the resulting analysis is mathematically
>% identical to the appropriate random effects (strictly called mixed
>% effects) analysis of these data.
|