Re: Statistica anova

Jerry Dallal (jerry@hnrc.tufts.edu)
Wed, 31 Aug 94 23:06:18 EDT

In article <778350911.AA05195@rochgte.fidonet.org>,
Paige.Miller@f313.n2613.z1.fidonet.org (Paige Miller) writes:
> Here's my data. I ran ANOVA through SAS, MINITAB, SYSTAT and STATISTICA.
>
> A B Y
> 1 1 1
> 1 1 2
> 1 1 1
> 1 2 3
> 1 2 2
> 1 2 3
> 1 2 2
> 2 1 9
> 2 1 8
> 2 1 7
>
> The table below shows the sums of squares from the four different packages.
> Package Type of SS SS due to A SS due to B
SAS 6.07 Type I 75.6000000 2.3333333
> Type II/III 66.6666667 2.3333333
SYSTAT 66.667 2.333
MINITAB 7.2 Seq SS 75.600 2.333
> Adj SS 66.667 2.333
STATISTICA Specific Effects 77.23189 11.26667

> Since the dataset is small enough, one can also compute the sums of squares
> "by hand" using formulas found in either Searle (1987) "Linear Models for
> Unbalanced Data" or Milliken and Johnson (1984) "Analysis of Messy Data".
> These hand calculations agree with SAS, SYSTAT and MINITAB.
>
> I can't explain STATISTICA's answers. Does anyone have an explanation?>

In fact, you *didn't* run ANOVA through SYSTAT. If you had run
cat a b
anova y
es
it would have failed due to the empty cell (2,2). What exactly did you do?
I would guess
model y=constant + a + b

To replicate Statistica's analysis in SYSTAT, use
cat a b
model y = a*b/means
es
hy
specify
a[1]=a[2]
test
hy
specify
b[1]=b[2]
test