If the aggregating hit rate for all of the Ganzfeld is considered in isolation, the existence of psi in the ganzfeld has been proven beyond any doubt. Consider a skeptic's 'conservative estimate' of the total hit rate (of which he personally tallied) of all post-1970 Ganzfeld: 28.6% over 6700 trials(SkepticReport.com). P(chance) = (1 - 0.999999999988614).
Unfortunately, the claims of methodological flaws as a cause of the abberations from chance still dominate some critic's minds, even if these these claims are absolutely baseless. What all camps acknowledge, including the fundamentalist skeptics, is that the methodological rigour of psi-setups has improved to the point where accusations of methodologic flaw hold no basis in reality. However, as for the Ganzfeld, researchers have not yet delivered the final nail into the denier's coffins. What, if possible, should be shown is that there exists either neutral (or positive, although there'd be no reason for this) correlation between continually increasing methodological rigour/reporting requirements and ES. Of course, some parapsychologists maintain that this is difficult due to a number of possible constraints. Unfortunately though, this has never been attempted for the entire database of the Ganzfeld. If and when this is done for the Ganzfeld, the Ganzfeld setup will have been epistemically, beyond all doubt, proven to be demonstrating Psi effects.
The reason I am commenting on this is because there exists an attempt at plotting an ES regression for the Ganzfeld database; yet, the database isn't a database, it's missing half of the data!
http://tinypic.com/view.php?pic=dq409u&s=5 (Bierman 2000)
There are numerous reasons why this graph doesn't do justice to the Ganzfeld data. Firstly, the pre-1986 data has been directly taken from Honorton's meta analysis (1985) of 28 Direct-Hit studies. These were markedly more successful (37% hit rate) than the Non-Direct-Hit studies, of which there were 14 studies. Whilst Honorton had perfectly sound reasons for excluding it from his meta-analysis, it causes an artifactual ES skew between 1970-1986.
Moreover, this ES regression assumes that Honorton's meta-analysis was overarching of the pre-autoganzfeld database, which it certainly was not. Storm & Ertel (2001) defined 11 Direct-Hit studies in between 1982-1986 that were excluded from Honorton's analysis; they called this database the 'S&E' database. 'The S&E database had an unweighted ES of 0.222 (SD = 0.23) and a Stouffer Z of 3.46 (p = 2.70 x 10E-4)' (Storm & Ertel 2001). Of course, this would have a significant effect on the supposed ES decline. I propose that if all Non-Direct-Hit data and this S&E data are accounted for in an appropriate way this data set may achieve an incline in ES leading to the end of the Old-Ganzfeld series, positively correlated to an inevitable increasing of methodological rigour.
Another concern with this regression is that it, whilst not in any fault of its author(s), discludes the successful post-MiltonWiseman experiments. These 10+ (highly significant) experiments will likely push the ES into pre-Hyman levels, essentially debunking the notion that methodological flaws are causing the success of the Ganzfeld.
It also seems that this regression was operating under the assumption that the Milton-Wiseman meta-analysis' exclusion criteria was sound. If not, you would have to wonder what exclusion/inclusion criteria Bierman used. Since it was not stated, I'd say it's safe to assume he stuck to Milton-Wiseman's critera and just included the experiments in that meta-analysis. This is a concern as this criteria was heavily criticised within the parapsychological community.
The graph should look more like this:
www.tinypic.com/view.php?pic=650a4x&s=5
As can be seen, the inclusion of the S&E and post-M&W databases would change the entire nature of the Ganzfeld ES regression. Further, I believe we must account for the non-direct-hit studies in the pre-joint-communique database in a statistically appropriate way, not only for statistical integrity, but to fully eliminate the charge of a diminishing ES. Also, I strongly believe that the above graph should be treated for the uniquely atypical musical target studies.
References:
http://skepticreport.com/sr/?p=316
http://www.parapsy.nl/uploads/w1/Benjaminschapter.pdf
http://www.aiprinc.org/para-ac02_Storm%20and%20Ertel_2001.pdf
Subscribe to:
Post Comments (Atom)
I think the major problem is that skeptics in general have convinced themselves that "meta analysis" is synonymous with "analysis with tons of potential holes," (even though countless other sciences use it all the time to no problem). And the problem with the entirety of the Ganzfeld database is that it's slightly antithetical to replication: there is a lot of variance in experimental methods in all of those trials, regardless.
ReplyDeleteWhat we need is a new paradigm, a test that isn't expensive or exhaustive or time consuming to run, that practically anybody could run, that shows consistent results under the exact same experimental conditions, and people run it the exact same way, every time. That's what Ganzfeld was supposed to be, and frankly IMHO exactly is, but too many people are bored of it.
Why don't you contact Andrew Endersby? I think he has the most complete version of the Ganzfeld Database of anyone. Run an analysis on the various types of experiments, and how each style correlates with ES. I think this might be what Endersby was trying to do, but got bogged down with.
If you do get your hands on it, please do share it with me though. I'll even help collaborate on a paper on the matter (I'm also running my own brand spankin' new experiments). But I think even if it were published, people would be a little loathe to look at it seriously, because it's like "Bleh, Ganzfeld, AGAIN?"
Interesting, I will think about contacting him for his database yet am weary of wasting his time as I lack statistical expertise. Although, creating a polynomial ES regression won't be too hard.
ReplyDeleteI doubt there will be an experiment that achieves positive ES every time, this is true of almost every area of research. Consider the 6 year aspirin study regarding heart health. The reason I think this is the case is because if psi does exist, instead of the results of each experiment oscillating around a baseline of 0 ES, it'll be oscillating around a baseline of y + x, where x = the true ES of psi. You'll still get negative ES, it's inevitable.
I have contacted skepticreport, waiting for their reply. Let's hope they deliver.
ReplyDeleteOkay so far I have accounted for all 50 post-manualganzfeld experiments, they are all the ones outlined in the autoganzfeld database and all the ones outlined in the milton/wiseman database and the 10 experiments that missed the M/W deadline but shortly proceeded it.
ReplyDeleteThere is a steep incline in the ES linear regression.
Now all that remains is account for the DH & non-DH manual ganzfeld databases and the E&S 'missing' 11 experiments between '82 and '86. I suspect we will end up with a neutral ES regression.
I have no access to the post-2000 data on the ganzfeld, if someone could fill me in I'd appreciate it.
I've got a good deal of statistical expertise; I'm taking the second semester of Experimental Methods and Statistics.
ReplyDeleteI have wanted to take a look at the entirety of the raw data, I just haven't had the time to track it down.
I have some ideas on how to tease out the actual effect size of the Ganzfeld. The trickiest part is that most of the tests are one run per subject, and so you get a lot of people making null-effect runs.