It's a good question, and one answer is: The minimally 100 post barrier, that ought to sort the wheat and the chaff, for the most part.
But yes, there are that sort of problems including selfreportation and privacy if filling out questionnaires. OTOH: It is interesting, in principle, to have a large basis of (presumed) patients with ME, with all manner of (non-)findings when they were tested, and all manner of experiences with all manner of medicines, and quite a few with at least one ill family-member a.s.o. There just is a lot of implicit information that may be interrelated in all manner of ways if only to find out about the population of the PR forums.
But again, as you said or suggested: reliability and privacy are important here, and I don't have all the answers at all, except that (1) once forms have been filled out one can anonimize them by way of hash-tables/systematic random numbering but (2) it seems to me the real problem who on the internet can link in to what as the forms are filled in.
Just thinking loud, again (and pretty F'd).
Here is some more about statistical possibilities, because I've been doing some searching: This looks quite interesting:
by a professor emeritus statistics including manuals for the program and for the statistics. This is the one I downloaded: It installs easily - with LOADS of statistics AND neural networks (for the affecionados ;)
Click here to download the free version OpenStatSetup.exe (INNO Setup).
Seems well done.
I found it at
where there is lots and lots more about statistics.