Date:         Fri, 14 May 2004 08:55:00 -0500
Reply-To:     Anthony Babinec <tbabinec@ameritech.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Anthony Babinec <tbabinec@ameritech.net>
Subject:      Re: implications of unequal sample sizes
Comments: To: "Nico Peruzzi, Ph.D." <nperuzzi@yahoo.com>
In-Reply-To:  <20040513215623.60999.qmail@web41502.mail.yahoo.com>
Content-Type: text/plain; charset="us-ascii"

Assuming that the cost of acquiring subjects is the same,
and that you are able to have the same number of
subjects per group, you get the most statistical efficiency
from having equal group sizes.  Extreme imbalance can
have implications for Type I and Type II error rates in
subsequent analysis, depending on the pattern of variance
within groups.

Anthony Babinec

-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
Nico Peruzzi, Ph.D.
Sent: Thursday, May 13, 2004 4:56 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: implications of unequal sample sizes

Thinking about two possible study designs and I'm looking
for some thoughts/ guidance on their implications.

Study 1: Subjects randomly assigned to 1 of 3 groups, in
equal numbers (e.g., say 200 in each group)

Study 2: Subjects randomly assigned to 1 of 3 groups, in
the following proportions 50%/25%/25% (e.g., 300 in group
1, and 150 in eachof groups 2 and 3).

Implications for:

-power?
-stats to use on the back end (assuming a numeric (ratio)
dependent variable)?
-other things I haven't thought of...?

Thanks in advance, Nico