Hi Daniel,
I thought I'd also throw in a few replies, and also summarize where I
think we've got to...
The main point that I am making, following Federico Turkheimer and the
Jernigan paper is:
Assertion 1) The thresholded map (e.g. significant areas for condition
A > condition B) is not useful for localization, because it can't tell
you if the activation was specific to these areas or not.
Assertion 2) The continuous map does a better job of localization than
the thresholded map (which does a very bad job), because the
continuous map shows you the level of activity across the whole brain
Then the more minor points:
Assertion 3) The continuous map does a much better job for meta-analysis.
I might be wrong, but I think we all basically agree on these points
in principle?
I'm going to try and summarize the objections - I hope I'm being fair here:
Objection 1) There's not much point in showing the continuous map,
because authors, reviewers and editors are well aware of the problem
(Assertion 1) already.
Objection 2) If we show the continuous map, we will have major new
problems with people overinterpreting them.
I don't know what to say about objection 1, because it seems to me
that I have rarely if ever seen an author trying to deal with
Assertion 1. It may be that most people in neuroimaging are checking
their activations with continuous maps, but that hasn't been my
experience. In any case, if we are interested in localization, then
surely we should present the evidence for localization in the paper?
For objection 2, I would only say that behavioral papers have been
reporting effect sizes for tests that did not pass significance
without causing chaos. I would much rather see someone making a good
argument for paying attention to trends in data, than using bad
statistics to make the data appear significant... But if this really
such a problem, then we should set ourselves the rule never to comment
on not-significant results, and we get the advantages without the
disadvantages don't we?
Thanks as ever for the nicely reasoned discussion...
Matthew
|