Dear Geoff,
> There have been a number of neuroimaging studies that have sought to
> address the question: is the neural activity evoked by cognitive process X
> in a location (dorsal/ventral/anterior/posterior) to that evoked by
> cognitive process Y? For example, a number of studies have examined frontal
> lobe responses with the aim of showing an inferior-superior separation of
> areas responsive to respectively object and spatial working memory.
>
> Different studies have used different statistical methods (or no method
> at all) to support claims of spatial segregation. I am curious what the
> readers of the list feel is the "state-of-the-art" method for these types of
> questions.
While I can hardly say what the best or most desirable method is, let me
just mention two different approaches that have been employed recently,
for the purposes of statistically testing differences between distribution
of foci in meta-analyses -
1) a Kolvogorov-Smirnov test, extended to 3 dimensions (Duncan & Owen,
2000, Trends in Neurosciences 23, 475-483), a non-parametric test which
in this paper was used to test for differences in distributions between
groups of foci.
2) a Hotelling's T-square test, a multivariate parametric T-test which we
used (Christoff & Gabrieli, 2000, Psychobiology 28, 168-186) to test for
differences in the mean/centroid location between two groups of foci (by
treating the foci as obsevations in multidimensional space).
I would be very interested to hear about other approaches that have been
used, or ideas about what can be used. The two approaches I've
listed above have their advantages and disadvanges. Hotelling's T-square
assumes a normal distribution, and if such is the case, it would be quite
sensitive to any differences. However, if the distirbution is bimodal or
something markedly else, Kolmogorov-Smirnov might be more approriate. On
the other hand, from what I know, Kolmogorov-Smirnov tends to be overly
conservative and it might fail to detect small differences.
Incidentlly, I happen to know that Ian Nimmo-Smith from MRC-CBU at
Cambridge has been interested and doing some work toward comparing
different approaches for statistical comparisons between groups of
3-dimensional foci. Perhaps he is reading this and would be willing to
offer his experience and advice?
Best wishes,
Kalina
_____________________________________________________________________________
Kalina Christoff Email: [log in to unmask]
Office: Rm.455; (650) 725-0797
Department of Psychology Home: (408) 245-2579
Jordan Hall, Main Quad Fax: (650) 725-5699
Stanford, CA 94305-2130 http://www-psych.stanford.edu/~kalina/
_____________________________________________________________________________
|