Hi

On 30 Mar 2004, John Katsaridas wrote:
> I am using Excel to do some ANOVA calculations. Say that we have one
> independent variable A, where we have 10 samples for each level of A.
> Then a one way ANOVA tells us whether there is a significant
> difference
> in the means of A's levels. 
> If I add a second independent variable B with 2 levels to the data
> set,
> and I assign 5 samples from each level of A to each level of B, then
> each cell of the 2 dimensional table will be made up of 5 samples.
> When I perform a two
> way ANOVA in Excel, it gives us 3 values, then means of the samples
> grouped
> by factor A, the means grouped by factor B, and the interaction
> effect.
> However the p value I get for the samples grouped by A in two way
> ANOVA,
> is different than the p value of the one way ANOVA on A that I
> performed
> at the beginning. Why is this? Shouldn't the p-values be identical?
> Aren't
> we testing the same hypothesis in each case?
> Thank you very much.

Your first anova partitioned SStotal into SS(A) and SS(SwA), with
the latter representing variability of Subjects within each
of the levels of A, and it being the basis for the denominator of
your F ratio.

Your second anova, which introduced B, partitioned SStotal into
SS(A), SS(B), SS(AxB), and SS(SwAB), with the last of these
becoming the error term.  

Although SS(A) is the same in both analyses, SS(SwA) will
generally be somewhat to much larger than SS(SwAB) and it will
have more degrees of freedom. So the two F ratios would normally
differ from each other, and could differ by quite a bit depending
on the amount of variability associated with B and AxB (i.e.,
the variation subtracted from SS(SwA).

Best wishes
Jim

============================================================================
James M. Clark                          (204) 786-9757
Department of Psychology                (204) 774-4134 Fax
University of Winnipeg                  4L05D
Winnipeg, Manitoba  R3B 2E9             [EMAIL PROTECTED]
CANADA                                  http://www.uwinnipeg.ca/~clark
============================================================================

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