I can't say. If it were my work, I would assume there's a hand-coding error, yet the errors are really too small for that, unless it's a very subtle one. On the other hand, I find it hard to believe that simply changing the version of matlab will lead to results that are that different.
The results really are very close. If you're satisfied with that, you could accept it and move on. If you're worried about the integrity of the results, you could try to pin down the difference. When I look for differences, I look in stages. E.g., are the endpoints of the preprocessing steps (say, the smoothed warped images) exactly the same? (Using e.g. unix command diff or cmp.) If not, is the difference extremely small? (E.g., 10^{-14}.) If the difference is extremely small, then you next go on to the statistics stage. One way to find a hand coding error in e.g. fMRI data (can't recall what yours is) is if the betas from the two cases are exactly the same for all runs but one. The odd run has the hand coding error.
Best,
Stephen J. Fromm, PhD
Contractor, NIMH/MAP
(301) 451--9265
________________________________
From: [log in to unmask] [[log in to unmask]]
Sent: Tuesday, September 08, 2009 10:09 PM
To: Fromm, Stephen (NIH/NIMH) [C]
Cc: [log in to unmask]
Subject: Re: RE: RE: question of the results
Hi Stephen:
do you know what causes the difference? AND which result should i follow?
bing
----- Original Message -----
From: "Fromm, Stephen (NIH/NIMH) [C]" <[log in to unmask]>
Date: Tuesday, September 8, 2009 12:59 pm
Subject: RE: RE: question of the results
To: "[log in to unmask]" <[log in to unmask]>
Cc: "[log in to unmask]" <[log in to unmask]>
> It's not huge, but it's more than I think it should be.
>
> From: [log in to unmask] [mailto:[log in to unmask]]
> Sent: Tuesday, September 08, 2009 12:49 PM
> To: Fromm, Stephen (NIH/NIMH) [C]
> Cc: [log in to unmask]
> Subject: Re: RE: question of the results
>
> Hi Stephen:
>
> the attached is the results from the two versions of matlab. Do
> you think the differece is huge?
>
> Could you let me know?
>
> THank you.
>
> bing
>
> Original Message -----
> From: "Fromm, Stephen (NIH/NIMH) [C]" <[log in to unmask]>
> Date: Friday, September 4, 2009 1:42 pm
> Subject: RE: question of the results
> To: "[log in to unmask]" <[log in to unmask]>
>
> > How little is "little"?
> >
> > Stephen J. Fromm, PhD
> > Contractor, NIMH/MAP
> > (301) 451--9265
> > ________________________________
> > From: [log in to unmask] [[log in to unmask]]
> > Sent: Friday, September 04, 2009 1:38 PM
> > To: Fromm, Stephen (NIH/NIMH) [C]
> > Subject: Re: question of the results
> >
> > Thank you Stephen. THe difference is little not substantial. DO
> > you think it is possible?
> >
> > thanks again.
> >
> > bing
> >
> > ----- Original Message -----
> > From: "Stephen J. Fromm" <[log in to unmask]>
> > Date: Friday, September 4, 2009 11:12 am
> > Subject: Re: question of the results
> > To: [log in to unmask], Bing Ye <[log in to unmask]>
> > Cc: "Stephen J. Fromm" <[log in to unmask]>
> >
> > > On Thu, 3 Sep 2009 10:38:39 -0400, Bing Ye
> > <[log in to unmask]> wrote:
> > >
> > > >Hi all:
> > > >
> > > >for spm 5 result part, would i get exactly the same result
> if i
> > > use two
> > > different versions of matlabs?
> > > >
> > > >I run spm5 from 7.1 version first and then run spm5 from 6.5
> > > version. but for
> > > the same subject, i got similar results not exactly the
> same, i
> > > wonder if the
> > > reason is the two different versions of the matlab.
> > >
> > > I'd be very surprised if there was any substantial difference,
> > > UNLESS there's
> > > some reason that SPM5 doesn't work properly with matlab
> > > 6.5. (But in that
> > > case, there should be an error and it shoudl stop running.)
> > >
> > > The _most_ I would expect would be a difference on the order of
> > > machine
> > > precision. Anything larger, something funny is going on,
> > > or maybe you made a
> > > trivial mistake entering something.
> > >
> > > >
> > > >Can anybody explain?
> > > >
> > > >thanks a bunch.
> > > >
> > > >bing
> > > >
> > >
> > >
> > >
> >
>
|