Re: [Marsbar-users] 4 way ANOVA - betas

[Gwladys R. <gw...@ex...>]

Dear Matthew=20

=20

Here is always my problem with my 4-way anova... I have a design with 3 =
factors, in 2 sessions, and 2 groups.

When I take the betas from single-subjects model and then perfom =
3-ANOVAs and 4-ways ANOVA I don't get the same results as Marsbar, when =
it performs 3-ANOVAs.

I saw in the forum your response about NaN value that could explain =
differences when we extract raw data from all subjects together or each =
subjects separately.=20

Is it the same explanation here or is there another one? or I made a =
mistake? Do you think I have to do another analysis ?=20

=20

Note : when I export  "summary.mat" from factorial design with 3 factors =
either for the two tasks separately or for the two groups separately, =
these two procedures don't give the same values (one condition one =
subject) nor the same results in ANOVAs (perhaps because of different =
scaling?) and these results are not exactly the same as those given by =
Marsbar.. I hope I am comprehensible.



Thank you for your attention and your long-suffering !



GR



----- Original Message -----=20
From: "Matthew Brett" <mat...@gm...>
To: "Gwladys Rey" <gw...@ex...>
Sent: Thursday, June 07, 2007 3:32 PM
Subject: Re: [Marsbar-users] 4 way ANOVA - export data


> Hi,
>=20
> Yes, you always have to estimate to get the betas.  The betas will not
> exist if you do not estimate...
>=20
> Matthew
>=20
> On 6/7/07, Gwladys Rey <gw...@ex...> wrote:
>> great thanks ! I'm confuse ; I have a doubt: I have to "estimate " =
with
>> Marsbar before getting the betas, or directly get the betas without
>> estimating ? it's not the same !
>>
>> ----- Original Message -----
>> From: "Matthew Brett" <mat...@gm...>
>> To: "Gwladys Rey" <gw...@ex...>
>> Sent: Thursday, June 07, 2007 1:43 PM
>> Subject: Re: [Marsbar-users] 4 way ANOVA - export data
>>
>>
>> > Hi,
>> >
>> >> All my subjects-model first level are estimated with SPM.
>> >> For one subject, I get the design, extract ROI, estimate. When I =
prompt E
>> >> =3D
>> >> mars_arm('get', 'est_design') then B =3D betas(E), I obtain 32 =
values in
>> >> the B
>> >> matrix.
>> >> Do these values (betas?) correspond to my 30 regressors for each =
subject
>> >> (and the 2 usual suplementary) : 8 conditions + 7 regressors of
>> >> non-interest
>> >> * 2 sessions ?
>> >
>> > Yes - exactly right.  So then it's just a question of getting the
>> > right betas out...
>> >
>> > Good luck,
>> >
>> > Matthew
>> >
>> >
>> >
>> > --
>> > No virus found in this incoming message.
>> > Checked by AVG Free Edition. Version: 7.5.472 / Virus Database:
>> > 269.8.11/837 - Release Date: 06/06/2007 14:03
>> >
>> >
>>
>>
>>
>=20
>=20
>=20
> --=20
> No virus found in this incoming message.
> Checked by AVG Free Edition.=20
> Version: 7.5.472 / Virus Database: 269.8.11/837 - Release Date: =
06/06/2007 14:03
>