Alpha is the proportion of times one would, WHEN THE NULL HYPOTHESIS IS TRUE, 
get data so extreme that one would reject the null hypothesis (falsely). That 
is what p=.05 represents. However, WHEN THE NULL HYPOTHESIS IS FALSE, alpha 
doesn't tell us anything at all. Beta (which we almost never even bother to 
measure, alas!) is the proportion of times we would get data so close to the 
parameter specified by the null that we would (falsely) fail to reject the null 
hypothesis. Power (1-beta) is the proportion of times we correctly reject the 
null (i.e, when the null is false).

So, it is impossible to tell what overall proportion of times the null is 
rejected (as a rather unreliable proxy for number of papers published) are 
cases of the null being *falsely* rejected. The two cells (power + alpha) don't 
"add up to one," so to speak, and since we don't know what (roughly speaking) 
the "base rate" is (i.e., the number of times we test true as opposed to false 
null hypotheses) we can't correctly weight them (power or alpha)  so that we 
can get the number we want.

Chris
---
Christopher D. Green
Department of Psychology
York University
Toronto, ON M3J 1P3
Canada

[email protected]
http://www.yorku.ca/christo/
==========================



On 2012-04-02, at 10:35 AM, Michael Burman wrote:

>  
> 
>  
> 
>  
> 
> I get a sense that I'm going to regret asking this question, but why wouldn't 
> setting the alpha value at .05 result in about 5% false positives in the 
> literature?  Are people suggesting that the true false positive rate would be 
> lower?  I get why it would be higher (statistical tricks, bias, research 
> short cuts, etc), but not why it would be lower.  
> 
> Mike
> 
> 
> -------
> Michael A Burman Ph.D.
> Assistant Professor 
> Dept. of Psychology 
> K-12 Outreach Coordinator for the Neurosciences
> University of New England
> 
> 328 Decary Hall
> 207-602-2301
> [email protected]
> 
> 
> ---
> 
> You are currently subscribed to tips as: [email protected].
> 
> To unsubscribe click here: 
> http://fsulist.frostburg.edu/u?id=430248.781165b5ef80a3cd2b14721caf62bd92&n=T&l=tips&o=17089
> 
> (It may be necessary to cut and paste the above URL if the line is broken)
> 
> or send a blank email to 
> leave-17089-430248.781165b5ef80a3cd2b14721caf62b...@fsulist.frostburg.edu
> 
> 
>  
> 
>  


---
You are currently subscribed to tips as: [email protected].
To unsubscribe click here: 
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=17094
or send a blank email to 
leave-17094-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu

Reply via email to