I knew you were going to read this thread.
ElectricAye
Posts: 4,561
I'm not going to make any comments on this one. Just be aware that it's going to be published by the American Psychological Association in their Journal of Personality and Social Psychology.
Maybe you can design your own experiments and try to replicate the effects?
Leon? Mike?
http://dbem.ws/FeelingFuture.pdf
Maybe you can design your own experiments and try to replicate the effects?
Leon? Mike?
http://dbem.ws/FeelingFuture.pdf
Comments
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1699970
Even Prof. Jahn, former leader of the PEAR people, who did a lot of work with subjects supposedly being able to influence the output of RNGs, was unable to replicate his findings in a different lab. That's where all these experiments fall down - no one seems able to replicate them, even the original experimenter.
If I was performing experiments of this nature, I'd assign p<.001, at least, as the acceptance criterion for a one-tailed test, given the importance of the findings. Bem seemed content with p<.05 for his first experiment, which means, crudely, that there is a 1/20 chance that the results could have been obtained by chance alone. Given the large number of subjects, which increases the power of a parametric test like the t-test he used, I can't see why he didn't use a higher level of significance.
There are no allegations of impropriety.
I appreciate your Halloween candor.
Now, I'm just browsing through the paper, and this is not a serious critique on my part, but I am getting the impression of a "divide and conquer" study of exactly that sort. There's a lot of complications to the design, and he still winds up with results that look like this (50% is the chance level):
53.9% - Hit rate on erotic pictures (significant)
49.6% - Neutral pictures (n.s.)
51.3% - Negative pictures (n.s.)
49.4% - Positive pictures (n.s.)
50.2% - Nonerotic romantic pictures (n.s.)
Combine this with the two failed replication attempts, and it looks like Type 1 error to me. Do we know how many times Bem ran the study to come up with these data? Nope. Do we have reason to suspect he ran it more than once? Yup. Read the caveats and the details of the design. He didn't come up with those things a priori: he ran failed studies and made changes in response.
Now, let's suppose for the sake of argument that the significant result reflects a real phenomenon. It would most certainly NOT be the same phenomenon that causes the lay public to believe in psi phenomena. I say that because a hit rate of 53% against a chance level of 50% would simply not be detectable other than through the use of statistics. In plain English, even if this were a real phenomenon, people would never notice it.
I will be impressed if this study can be routinely replicated by people regardless of their stance on the question. But even then I see no reason to associate the findings with routine layperson's claims of psi. Bem's results don't bear on those claims.
'
I see why caly didn't pass the Green Bud Bill.(Making Marijuana legal)
'
Time to buy some stock in Oreo cookies. (their gonna have the munches-sales are up)
You're correct of course! :smhair:
Mike B.
http://www.randi.org/site/index.php/1m-challenge.html