MC> RESPONSE OF STATSOFT, INC.
MC> TO MS. PAIGE MILLER
That should be Mr. Paige Miller.
MC> 3. When the user requests to compute the "A main
MC> effect," the program will first issue a message,
MC> LEAVING NO AMBIGUITY concerning what specific
MC> hypothesis regarding a linear combination of means
MC> will be tested. Specifically, the message reads:
MC> Because of missing cells, the following specific
MC> hypothesis will be tested:
MC> Factor A B Contrast
MC> 1 1 0.5
MC> 1 2 0.5
MC> 2 1 -1.0
Different statisticians will prefer to test different contrasts. I happen to
feel that this contrast is meaningless when there is a missing cell. The only
fair comparison for factor A is
A B Contrast
1 1 1
1 2 0
2 1 -1
which is what SAS, SPSS, Minitab and others give you as a default. When you
include the data from A=1, B=2 as Statistica does, and there is a significant
effect of factor B, then you have an unfair comparison.
MC> 5. The issue of Type III and IV SS in incomplete
MC> designs is of course also discussed in great detail in
MC> the literature. For example, Milliken and Johnson,
MC> Volume I, 1984 (Chapter 14) show how in incomplete
MC> designs the standard Type III SS will actually test
MC> complex compound hypotheses about means and Cell N's.
MC> 5.1. Here is another simple example; try analyzing
MC> these data:
MC> A B Y
MC> 1 1 1
MC> 1 1 2
MC> 1 2 3
MC> 1 2 4
MC> 2 1 5
MC> 2 1 6
MC> The default results computed by SPSS are:
MC> Sum of Mean Sig
MC> Source Squares DF Square F of F
MC> Main Effects 16.000 2 8.000 16.000 .025
MC> A 16.000 1 16.000 32.000 .011
MC> B 4.000 1 4.000 8.000 .066
MC> Is there really a marginally significant B main
MC> effect?
Yes indeed, there is. The STATISTICA contrast ignores the fact that the main
effect of A has had on the data. The only fair comparison due to the missing
cell is to do a comparison of the levels of B while the level of A is held
fixed.
Paige Miller
Paige.Miller@f313.n2613.z1.fidonet.org or kp40.118980@kodako.kodak.com ___
Blue Wave/QWK v2.12