On Mon, 15 May 2006 14:15:11 -0700, Stuart McKelvie wrote:
>
>Dear Stephen,
>
>You can run this with unequal n. Look at variances to see how
>much they violate the homogeneity assumption.

Just to emphasize the above points a little more, Glass and Hopkins
(Statistical Methods in Education and Psychology, 3rd Ed) review
the issue of unequal sample sizes and unequal variances for the
two group situation (see Chapter 12, specifically pp290-295).
The ideas extend to the mult-group ANOVA situation.  In the
two group situation, Glass & Hopkins suggest the Welch t-test
(aka seperate variance t-test; only the calculation of the degrees
of freedom are affected).  The old statistical package BMDP used
to calculate an Welch-type ANOVA for the multi-group situation
and contemporary statistical packages (SAS' GLM procedure)
do the same.

>Another idea: Choose a random set of 15 cases from
>the normal group. Check it their mean matches the mean of
>the total normals. If so, then run an ANOVA with 11, 15 and 17.

An alternative analysis would be use all of the data and to do
a randomization test on the three groups but you'd probably want
to consult with someone familiar with these types of tests and
the software that implements them (though SAS does allow one
to do so; see
http://64.233.179.104/search?q=cache:m88pa-siEnAJ:www2.sas.com/proceedings/sugi27/p251-27.pdf+%22randomization+test%22+ANOVA&hl=en&gl=us&ct=clnk&cd=1
or
http://tinyurl.com/mx3dx

-Mike Palij
New York University
[EMAIL PROTECTED]

>Stuart
>Stuart J. McKelvie, Ph.D.,
>Department of Psychology,
>Bishop's University,
>Route 108 East,
>Borough of Lennoxville, Sherbrooke,
>Qu¨bec J1M 1Z7,
>Canada.



---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

Reply via email to