On Wed, 25 Jun 2003, John Mercer wrote:

>  [EMAIL PROTECTED] (Karl L. Wuensch) wrote:
>
> > Dunnett's test ... each of several means is contrasted with a single
> > reference mean.  ...
>
> So would this be the proper choice if I have a single experimental
> group and three control groups?

Depends partly on why you have three control groups.  If they were
thought to be equivalent, or exchangeable, it would probably be
efficient to pool them into a single group.  If they are not (e.g., if
each control group is "controlling" for a different thing), will you be
interested in differences between the several controls?  (If so, this
would argue against Dunnett's test, since not all the comparisons of
interest are of that kind.  But if you are indifferent to whatever
differences may exist between control groups, and pairwise comparisons
involving your one experimental group are your ONLY interest, Dunnett's
would seem appropriate.)

> I am in this situation, with very large standard deviations caused
> by uncontrollable factors. . .

Does that mean your control groups aren't controlling for much of
anything?  You don't describe the patterns of means OR s.d.s in your
several groups.  If the control groups have large s.d.s but not the
experimental group, and if the control group means appear more similar
than the (E-C) differences, I'd be inclined either to pool the control
groups or to construct a contrast of the form (3*E - C1 - C2 - C3), and
use the Scheffe method (since the choice of contrasts would have arisen
from consulting the results, hence a thoroughly post hoc method of
making comparisons is required).  Of course, if your s.d.s are REALLY
large, you probably can't tell anything anyway...
  HTH.   -- DFB.
 -----------------------------------------------------------------------
 Donald F. Burrill                                         [EMAIL PROTECTED]
 56 Sebbins Pond Drive, Bedford, NH 03110                 (603) 626-0816

.
.
=================================================================
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