Print

Print

Dear Donald and Helmut,

 

thanks again for your advices.

Now that I’ve tried it:

Flexible factorial does make a difference in repeated measures.

 

I’ve read your previous mails in the last couple of years and it seems that I am not the first who tried to analyze mixed models this way…

 

Still I have a maybe naïve question:

I am still interested in the effect of group (original design 2 groups, 2 factors (2 and 3 levels, respectively = 6 conditions per subject).

 

I am afraid that just averaging the images over the 6 conditions and entering them in an ANOVA for a group comparison will dramatically reduce the power.

 

Is it “allowed” to redesign the flexible factorial analysis for the sake of the group analysis and just enter the factors “subject” and “group” but none of the within-subject factors or interactions?

This would preserve the 6 observations per subject.

Possibly that’s a bit too creative but can I get around the error term problem this way?

 

Thanks in advance

 

Juraj

 

 

 

 

Von: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] Im Auftrag von MCLAREN, Donald
Gesendet: Montag, 20. Januar 2014 20:46
An: [log in to unmask]
Betreff: Re: [SPM] Biological Parametric Mapping and Full Factorial Design

 

 

Dear Donald,

 

many thanks for your answer.

However, I am pretty sure that full factorial can be used as well:

you just determine the between group factor levels to be independent and the within group factor levels to be dependent.

From my understanding, the reason for using flexible factorial is if not all main effects and interactions are to be tested.

Please correct me if I’m wrong.

 

No. This is incorrect. The flexible factorial needs to be used to include the subject factor in the repeated-measures analysis. Furthermore, you can only test within-subject effects in the models as the model only contains the within-subject error term.

 

 

@ 2) I do believe that in the case of my study, including VBM data as covariate does make sense as they may influence the main effect of group and group by within-subject factor interactions.

 

You can't evaluate the main effect of group in a repeated-measures design as the error term is wrong. What you are suggesting is a three-way interaction with your covariate. The easiest way to do this, that will be statistically correct is to create difference images of your DV or compute the slope of your DV. Then test whether the slope is correlated with VBM values. For the effect of group. Use the average value across your DV.

 

 

So: any ideas how to implement BPM?

 

See above. Simplify the analysis and removed the repeated element. 

 

Thanks,

 

Best regards

 

Juraj

 

 

Von: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] Im Auftrag von MCLAREN, Donald
Gesendet: Montag, 20. Januar 2014 17:17
An: [log in to unmask]
Betreff: Re: [SPM] Biological Parametric Mapping and Full Factorial Design

 

Two things:

(1) If you have 2 groups and 2 within-subject factors, you can only look at within-subject factors AND this should be done with the flexible factorial. The full factorial is not correct for repeated-measures.

 

(2) Covariates should not be added to the model as they do not change the within-subject effects.

 

I'm not sure if they are allowed in BPM, but if they are they will likely lead to the wrong results as with SPM.


Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
intended only for the use of the individual or entity named above. If the
reader of the e-mail is not the intended recipient or the employee or agent
responsible for delivering it to the intended recipient, you are hereby
notified that you are in possession of confidential and privileged
information. Any unauthorized use, disclosure, copying or the taking of any
action in reliance on the contents of this information is strictly
prohibited and may be unlawful. If you have received this e-mail
unintentionally, please immediately notify the sender via telephone at (773)
406-2464 or email.

 

On Mon, Jan 20, 2014 at 8:22 AM, PD Dr. med. Juraj Kukolja <[log in to unmask]> wrote:

Dear community,

 

I wonder whether there is a possibility (script or trick) to perform a full factorial analysis (with between-group and within-group factors) with BPM.

I would like to include VBM data as a covariate in an fMRI analysis (2 groups, 2 within-subject factors (with 2 and 3 levels, respectively)).

 

So far, BPM only permits between-group tests, if I am correct.

 

Many thanks in advance!

 

Best regards

 

Juraj