Zoltan Nagy wrote:
> Dear Will and Seyed,
> On the same subject may I ask what is meant by "prior hypotheses" here? For
> example, suppose
>
> (1) you run the images of a group of subjects and you find an activation at
> x, y, z.
> (2) then on a closer look at your original data you realize that in your
> covariate measurements (i.e. the task score) you have an outlier
> (3) then you redo the analysis without including the subject who was an
> outlier on the cognitive score
>
> now, you know exactly where you want the activation (from step one above you
> got an activation and you would like to have it in the same place). but is
> this really a prior hypothesis? in other words: does one you need to use the
> corrected p-values or not? i am asking because i am in a discussion with
> someone about this question. i would personally think that this would be a
> false approach since you made your hypothesis on a finding which may have
> been corrupted by the outlier. but then again i might be wrong and in
> practice it works out fine.
>
By `prior' i mean prior to collecting the data.
So I would say in this case you do not have a prior hypothesis.
So you would use corrected p-values.
Best wishes,
Will.
> Thanks in advance for comments (references to books or articles are also
> welcome),
> Zoltan
>
> **********************************
> Zoltan Nagy
> Karolinska Institutet
> Neonatology
> Astrid Lindgrens Barnsjukhus Q2:07
> 171 76 Stockholm, Sweden
>
> Phone: +46-8-517-77354
> Fax: +46-8-517-77353
> E-mail: [log in to unmask]
> **********************************
>
>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7478
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|