Hacker News Reader

Actually the paper discussed in OP is a preprint at: http://arxiv.org/abs/1006.4848

It uses an extremely suspicious technique to test the significance of their finding. I doubt this paper will pass review without substantial changes.

edit: The one you posted is 2 years old, but tries to explain the same effect with a different cause.

View on HN

For instance, eq. (7) used to calculate significance values is the invention of the first author and is not widely accepted. Also, eq. (1) assumes joint independence of all M = 22 frequency bands, which has not been justified.

Can you elaborate on the suspiciousness for a lay audience?

I'm referring here to the significance estimates of section 3.

It's immediately suspicious that the authors introduce a new type of "test" for whatever spectra are being measured. When that happens, the authors need to rigorously demonstrate its validity. That is not present in this preprint.

The thing that needs to be demonstrated is that their method generates a reasonable space of outcomes. It's not clear to me that's true for their method.

Reply on news.ycombinator.com