sure. my understanding is that mixed effects OLS ignores lower level
variances. why do you not recommend 1st stage FLAME, which is quoted to be
just as fast as OLS, but would allow me to specify separate groups for
variance calculation and has more accurate variance estimation (perhaps)?
vitaly
On Thu, 12 Jan 2006, Steve Smith wrote:
> Hi - the multiple messages must be because FLAME is running one slice at a
> time to save memory - it does this as soon as you have a large number of
> input subjects. If you have a large number the difference between full FLAME
> and OLS is small, so you could just run ME OLS.
>
> Cheers. Steve.
>
>
> On 11 Jan 2006, at 22:19, Vitaly Napadow wrote:
>
>> hi.
>>
>> i just meant that when i flipped the var/copes i renamed them _f, so when
>> feat looked for tdof*_f it wasnt there.... but i solved that (thanks!), and
>> am running my analysis, which is taking *days* for full flame - i have 68
>> inputs, multiple contrasts, no f-tests.
>>
>> when i break the 2x2x2 anova analysis in two - e.g. 2x2 anova, the number
>> of inputs is 37, and report.log says analysis goes straight to directory
>> "stats":
>>
>> Log directory is: stats
>> Setting up:
>> No f contrasts
>> nevs=4
>> ntpts=35
>> ngs=1
>> nvoxels=214950
>> Running:
>> nmaskvoxels=44309
>> njumps = 10000
>> burnin = 500
>> sampleevery = 1
>> nsamples = 9500
>>
>> Metropolis Hasting Sampling
>> Number of voxels=44309
>> Percentage done:
>> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
>>
>> ....
>>
>> however, for the 68 input analysis that is taking forever, report.log shows
>> iteration after iteration varying "Number of voxels", such as:
>>
>> /usr/pubsw/packages/fsl/3.2b/bin/flame --cope=tmpcope --vc=tmpvarcope
>> --dvc=tmptdof --mask=tmpmask --ld=stats31 --dm=design.mat --cs=design.grp
>> --tc=design.con --fc=design.fts --ols --nj=10000 --bi=500 --se=1 --fm
>> --zlt=2.25 --zut=2.65
>> Log directory is: stats31
>> Setting up:
>> nevs=8
>> ntpts=68
>> ngs=1
>> nvoxels=4784
>> Running:
>> nmaskvoxels=345
>> njumps = 10000
>> burnin = 500
>> sampleevery = 1
>> nsamples = 9500
>>
>> Metropolis Hasting Sampling
>> Number of voxels=345
>> Percentage done:
>> 1 2 3 4 5 6 7 8 9 10 11 12
>>
>> ...
>>
>>
>> can you tell from these snipets if there could there be a problem with my
>> input copes? or do i need to upload these different directories for you to
>> be able to tell why the longer analysis has these multiple stats31, 32, 33
>> etc. steps.
>>
>> thanks again - the basic setup is a 2x2x2 anova - extention of the example
>> in the manual for a 2x2 anova...
>>
>> -vitaly
>
>
> ------------------------------------------------------------------------
> ------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
>
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
> ------------------------------------------------------------------------
> ------------------
>
>
>
--
|`~~~~~~~~~~~~~\|/~~~~~~~~~~~~~'|
Vitaly Napadow, Ph.D., Lic.Ac.
Dept. of Radiology
Massachusetts General Hospital
CNY 149-2301, 13th St.
Charlestown, MA 02129
phone: (617) 724-3402
fax: (617) 726-7422
email: [log in to unmask]
|`~~~~~~~~~~~~~/|\~~~~~~~~~~~~~'|
|