On 27 Nov 2002 13:17:52 -0800, [EMAIL PROTECTED] (dennis roberts) wrote:

> 
> >
> > >
> > > (I don't understand it, but from what I have read I believe the usual
> > > technique of simply adding the squares of the SDs is not appropriate
> > > because I have (hopefully) correlated data, and different sample
> > > sizes)
> > >
> 
> this has me puzzled ... how can there be a dependency IN the data IF you 
> have different samples of different sizes? 
> 

Dennis has pointed to a fatal *flaw*  in your question.

Do you mean, perhaps, that the groups have means that
are extremely different?

 - in that case (which does, after all, fit one notion that I have 
of  'correlated data')  it would be a bad idea, for most people in
most contexts, to present the overall mean and its overall
standard deviation.  For instance you could have different samples
from different people;  then you might want to estimate the 
variance "within-person", or  "between person", or both.

Those could be useful, for some purposes.  Hardly ever 
would someone want to see the arbitrary combination of
multiple, uncontrolled samples.  (Unless there is some
future use that will use exactly the same Ns....)

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to