[Air-L] A question for researchers interested in the basics of statistical inference

[Peter Timusk]

Hi I missed this when it was first asked but I just read the two (I  
think that's the number) replies to the list and both seem well  
informed to me. I wish I could write as much and as well about  
statistics even though I understand both replies. =;+)


I will add this. That also 900 or so units of analysis should provide  
a good normal distribution. I use magic numbers for samples like this  
12, 20, 30, 50, 100, 300. 12 is enough for small sample T test and as  
you go to larger samples you get closer to a normal distribution and  
more tests become justified.

Also Fred's reply got me thinking that you could check with linear  
regression diagnostics ( See a book on this. Neuter et al  Linear  
Models is my bible)  to see if the needed assumptions (Fred listed  
these) hold for linear regression.

Also I have studied in drug law about survey's of large populations  
like returning Vietnam veterans and national health survey's so I do  
not believe that you have it right about there being no survey's with  
probabilistic sampling. May be I am wrong but you should check in the  
drug law literature to verify your statement. What is interesting in  
these survey's is that we find most addicts recover and become drug  
free, yet if you ask hospitals they do not say that most addicts  
recover. This is also the same for mental illness. I think it has to  
do with the hospitals only seeing the problem cases thus as observers  
they are biased.

I would also think those who use the Internet and openly talk about  
illegal drugs online are different from those addicts who know they  
will be monitored by posting online and do not post. This suggests if  
true that like Fred was saying your sample would have bias and you  
would have to qualify your sample with some statement about the units  
of analysis opening up online. The population at large may not open up  
online.

Also in self help groups confidentiality is practiced so nothing would  
tend to show up online about these meetings.

Peter president of an unmentioned self help group at present.

On 6-Sep-09, at 7:32 AM, Monica Barratt wrote:

> Thanks to everyone who has responded, publicly and privately. I'll  
> get back
> to this thread over the next week, as I expect to get further input  
> from a
> few other contacts soon too.
>
> Monica
> _______________________________________________
> The Air-L at listserv.aoir.org mailing list
> is provided by the Association of Internet Researchers http://aoir.org
> Subscribe, change options or unsubscribe at: http://listserv.aoir.org/listinfo.cgi/air-l-aoir.org
>
> Join the Association of Internet Researchers:
> http://www.aoir.org/
>

Peter Timusk statistical computer programmer
ptimusk at sympatico.ca
address 701-151 Parkdale Avenue
Ottawa, Ontario Canada K1Y 4V8
Phone 613-729-8328

May all your numbers be quality numbers... even if they are only  
average numbers.