Hi all,
This very entertaining thread reminds me that I never announced the
existence of this mailing list to several software-centric mailing lists:
neuro-mult-comp
http://brainvis.wustl.edu/mailman/listinfo/neuro-mult-comp
Rather than cross-post to multiple tool-centric lists (e.g., AFNI,
caret-users, Freesurfer, FSL, SPM), neuroimaging researchers can discuss
multiple comparisons and thresholding problems and issues on this
algorithm-centric list.
I realize what starts out software-centric easily morphs into a more
algorithmic-centered discusssion, as this thread certainly illustrates:
In the case of this particular thread, Lauren Parks' initial, innocent
query appeared SPM-specific, but the ensuing discussion is most welcome
and appropriate on this new list.
If you are cross-listed, you may see duplicate announcements.
Donna Hanlon
On 02/09/2006 05:08 AM, Thilo Kellermann wrote:
>Hi All,
>
>without solving the problem I just like to cite Gerd Gigerenzer, who wrote a
>quite interesting book chapter on the topic of statistical reasoning, which
>is available online:
>http://www.mpib-berlin.mpg.de/dok/full/gg/ggstehfda/ggstehfda.html
>
>"The superego, the ego, and the id in statistical reasoning"
>In: "A handbook for data analysis in the behavioral sciences"
>Editors: Keren, G. and Lewis, C.
>Hillsdale, NJ: Erlbaum, 1993 (pp. 311-339)
>
>have fun reading the following paragraphs coming from the above mentioned
>chapter,
>
>Thilo
>
>"Freudian analogy":
>
>"The Neyman-Pearson logic of hypothesis testing functions as the Superego of
>the hybrid logic. It demands the specification of precise alternative
>hypotheses, significance levels, and power in advance to calculate the sample
>size necessary, and it teaches the doctrine of repeated random sampling. The
>frequentist Superego forbids epistemic statements about particular outcomes
>or intervals, and it outlaws the interpretation of levels of significance as
>the degree of confidence that a particular hypothesis is true or false.
>
>The Fisherian theory of significance testing functions as the Ego. The Ego
>gets things done in the laboratory and gets papers published. The Ego
>determines the level of significance after the experiment, and it does not
>specify power nor calculate the sample size necessary. The Ego avoids precise
>predictions from its research hypothesis; that is, it does not specify the
>exact predictions of the alternative hypothesis, but claims support for it by
>rejecting a null hypothesis. The Ego makes abundant epistemic statements
>about particular results. But it is left with feelings of guilt and shame for
>having violated the rules.
>
>Censored by both the frequentist Superego and the pragmatic Ego are statements
>about probabilities of hypotheses given data. These form the Bayesian Id of
>the hybrid logic. Some direct measure of the validity of the hypotheses under
>question - quantitatively or qualitatively - is, after all, what researchers
>really want."
>
>(written by Gerd Gigerenzer)
>
>
>
|