Robert Herdegen asked about what kind of error bar is to be prefered, 
and Mike Scoles said:

> I don't know if there is a convention, but 1 standard error has always
> made sense to me.  If the error bars for two groups do not overlap,
> you can be reasonably confident (but not certain) that the means are
> significantly different.

Nope. It's three error bars between the means, not two as Mike 
suggests and as most think.  See Dunlap and May (1989). 

I like standard errors best, for two reasons:

1) shorter than other choices, hence less clutter on the graph

2) lend themselves well to eyeballing whether significant or not as 
above. In practice, I note the gap between the opposing error bars of 
the two groups, and estimate whether an average of the two will fit 
between. If yes, it will probably be significant at p = .05. Of 
course, the more the gap exceeds this minimum, the more it is likely.

Stephen

Dunlap, W. and May, J. (1988), Judging statistical significance by 
inspection of standard error bars. Bulletin of the Psychonomic 
Society, 27, 67-68.
___________________________________________________
Stephen L. Black, Ph.D.            tel:  (819) 822-9600 ext 2470
Department of Psychology         fax:  (819) 822-9661
Bishop's  University           e-mail: [EMAIL PROTECTED]
Lennoxville, QC  J1M 1Z7
Canada

Dept web page at http://www.ubishops.ca/ccc/div/soc/psy
TIPS discussion list for psychology teachers at
 http://faculty.frostburg.edu/psyc/southerly/tips/index.htm    
_______________________________________________


---
You are currently subscribed to tips as: [EMAIL PROTECTED]
To unsubscribe send a blank email to [EMAIL PROTECTED]

Reply via email to