(2019-07-29, 10:14 AM)Chris Wrote: Unfortunately that's only the first 8 minutes. The rest doesn't seem to be available online.
Someone kindly sent me a private message saying that there was more information about David Mandell in the thesis of Chris French's former student, Louie Savva, which is available here:
https://pdfs.semanticscholar.org/41be/a5...e6d96e.pdf
It seems Savva started parapsychology research expecting to find experimental confirmation, and did research for a Ph.D. supervised by French, and afterwards spent several years as a postdoctoral researcher, but ended up disillusioned with the field and became quite a vocal critic. Some more information about him can be found in this mind-energy.net discussion from 2008:
https://forum.mind-energy.net/forum/skep...psychology
An archived copy of Savva's explanation can be found here. He says he wasn't going to finish writing up his thesis, but was finally motivated to do so "in the hope that I might dissuade others who are interested in the question of the paranormal, from pursuing it any further." (From internal evidence, it doesn't actually seem to have been finished until 2014.)
https://web.archive.org/web/201902241431...ology.html
Anyhow, details of Savva's investigation of David Mandell can be found in Chapter 5 and Appendix B. It was the statistical analysis that particularly intrigued me. The procedure was 40 of Mandell's "best" pictures were provided, together with details of the events that he believed they related to. Then Savva searched news archives and found 40 alternative events which also resembled the pictures. These were then submitted to judges (with some attempt at blinding, though the circumstances partly defeated it), who were asked to rate the similarity of the pictures to both Mandell's events and the alternative events. Their responses were analysed to see whether the similarity ratings for Mandell's events were significantly higher.
I don't think it's at all easy to come up with a valid statistical test for data like these, but from the test that was actually done I think it would be hard to conclude anything, whatever the results. Obviously a lack of statistical significance doesn't necessarily mean there's no difference. It can just mean your procedure doesn't have enough statistical power. There's a particular difficulty in this case. While in some cases Mandell's events came several years after the paintings, on average they were reasonably close (range 1 day to 8 years, median 12 months). But in contrast, Savva allowed himself a much longer period to choose his alternative events from. About three quarters of the alternative events were later than the paintings, and for those the range was 14 months to 10 years, median 6-7 years. About a quarter were earlier than the paintings, with range 20 months to 52 years, median 12 years.
So even if the comparison had shown no significant difference between the similarity to the paintings of Mandell's events and Savva's alternative events, I think it would have been difficult to conclude anything. But anyway, that wasn't the case. Of the 40 comparisons, 31 were rated more similar to Mandell's events, with the difference being statistically significant (p<0.05) in 7 cases, and extremely small p values in some cases (0.0001 and 0.0002 for the two pictures interpreted as showing the Twin Towers). Nevertheless, after some further discussion, Savva's conclusion was that Mandell did not have precognitive dreams of the future. Savva thought the matching of targets was the result of "subjective validation" on Mandell's part.
My own feeling is that the only basis for an objective judgment about Mandell's pictures would be a valid statistical test, fixed in advance. Without that, any opinion is bound to be subjective.
(Edit: I meant to add that Louie Savva's thesis is 234 pages long and reports on several different studies, so there is plenty of other potentially interesting stuff in there. Also, for people who would like a closer look at some of David Mandell's pictures, the seven pictures for which the Mandell events were rated significantly cmore similar than the alternative events, are reproduced in the appendix.)