Strange design output for what should be straightforward...

[Ben Beardmore]

Here is a straight forward orthogonal design in 32 rows. All 6 attributes are specified to have 4 levels. I've pasted the design below, and for some strange reason, attribute one only has 3 levels represented. What's going on?!

Design
;alts = alt1, alt2
;rows = 32
;orth = sim
;model:
U(alt1) =
b1 +
b2 * A[0,1,2,3] +
b3 * B[0,1,2,3] +
b4 * C[0,1,2,3] +
b5 * D[0,1,2,3] +
b6 * E[0,1,2,3] +
b7 * F[0,1,2,3]
$

Choice situation alt1.a alt1.b alt1.c alt1.d alt1.e alt1.f
1 2 2 0 3 1 0
2 2 0 0 0 0 2
3 1 1 0 1 0 0
4 1 3 0 1 2 1
5 2 0 0 3 3 1
6 2 2 0 0 2 3
7 0 3 0 2 1 2
8 0 1 0 2 3 3
9 1 2 2 2 0 3
10 0 2 2 1 3 0
11 2 1 2 3 2 0
12 2 3 2 3 0 1
13 2 1 2 0 1 3
14 1 0 2 2 2 2
15 2 3 2 0 3 2
16 0 0 2 1 1 1
17 1 3 1 2 3 0
18 2 0 1 3 1 3
19 0 1 1 1 2 2
20 1 1 1 2 1 1
21 2 0 1 0 2 0
22 2 2 1 3 3 2
23 0 3 1 1 0 3
24 2 2 1 0 0 1
25 2 3 3 3 2 3
26 1 2 3 1 1 2
27 1 0 3 1 3 3
28 0 2 3 2 2 1
29 2 1 3 0 3 1
30 2 3 3 0 1 0
31 2 1 3 3 0 2
32 0 0 3 2 0 0

[admin]

Hi Ben,

Strange indeed. I am looking into this closely, but in the meantime you could just add another superfluous attribute. The problem does not appear in this case, and the extra column can just be ignored - all other attributes will be orthogonal. See the syntax below.

Cheers,
Andrew

Code: Select all

Design
;alts = alt1, alt2
;rows = 32
;orth = sim
;model:
U(alt1) =
b1 +
b2 * A[0,1,2,3] +
b3 * B[0,1,2,3] +
b4 * C[0,1,2,3] +
b5 * D[0,1,2,3] +
b6 * E[0,1,2,3] +
b7 * F[0,1,2,3] +
b8 * G[0,1,2,3]
$

[Ben Beardmore]

That's the solution that I eventually used. I just thought I should bring it to your attention, as it seems like a very strange bug.

[Ben Beardmore]

With the update to 1.0.2 this now seems to have been resolved. Thanks!

[Andrew Collins]

Yes we addressed this problem with 1.0.2. The problem stemmed from the underlying orthogonal tables. The columns of the table can be combined to great columns with more levels - for example two columns with two levels each can be combined to create a column with four levels. With some tables this property does not hold, and the new column will introduce correlations or not have the correct number of levels. We have introduced a check to prevent these cases.

Andrew