OOD choice sets repetition

[Mariamishaq]

Hello Prof and admin members,
I ran the following syntax with OOD(optimal orthogonal design) and got 90.86% D optimality. However, I have a query in the choice sets. Each block has at least one choice set that is repeated in the preceding set. Is that normal and for a specific purpose or is this a problem that I should resolve?

Design
;alts= optA*, optB*, Neither
;rows=36
;orth=ood
;block=4
;model:
U(optA) = b1 [0] * Price[7.75, 10.5, 13.25, 16]
+ b2.dummy[0]* Carbonfriendly[1,0]
+ b3.dummy[0]* Humanelyraised[1, 0]
+ b4.dummy[0] * Traceability[1, 0] /
U(optB) = b1 * Price
+ b2 * Carbonfriendly
+ b3 * Humanelyraised
+ b4 * Traceability
$

Waiting for your response.
Thanking you in anticipation,
Mariam Ishaq

[Michiel Bliemer]

Hi Mariam,

This issue seems to arise because there does not exist an orthogonal design with 36 rows, so Ngene creates in this case a near-orthogonal design and in this case looks like something that I would not use.

You need to increase the number of rows to 40, for which an orthogonal design exists. Have a look at the design with 40 rows and see if you spot any further issues.

Michiel

[Mariamishaq]

Thank you Prof for your response. With 40 choice sets the repetition is reduced to one in block 4. Is that ok or should I try something else?

Best,
Mariam Ishaq

[Mariamishaq]

Hello Prof and admin members,
In addition to the problem of having a repetitive choice set in block 4 after making the rows 40 , I am encountering another issue: the choice with lesser price shows all the attributes i.e carbon friendly, traceability and humanely raised while the one with much higher price shows none of the attributes in two of the choice sets. This seems like an awkward choice to give to a respondent. Is it common in Optimal orthogonal or this needs to be addressed?

Waiting for your response,
Mariam Ishaq

[Michiel Bliemer]

It is common for any orthogonal design to have choice tasks with dominant alternatives or attribute level combinations that may not make sense because an orthogonal design ignores the meaning of the attribute levels. The only way to avoid such problematic choice tasks is to move away from orthogonal designs and impose dominance constraints and/or other constraints on the design.

For example, the syntax below would avoid repetitions and dominant alternatives (assuming that I set the sign for each prior correctly). Note that you have more flexibility in selecting the number of rows and blocks than you have with orthogonal designs, so you could use 24 rows but 36 rows is also fine.

Code: Select all

Design
;alts= optA*, optB*, Neither
;rows=24
;block=3
;eff = (mnl,d)
;alg = mfederov
;model:
U(optA) = b1[-.00001] * Price[7.75, 10.5, 13.25, 16](5-7,5-7,5-7,5-7)
+ b2.dummy[0.0001]* Carbonfriendly[1,0]
+ b3.dummy[0.0001]* Humanelyraised[1, 0]
+ b4.dummy[0.0001] * Traceability[1, 0] /
U(optB) = b1 * Price
+ b2 * Carbonfriendly
+ b3 * Humanelyraised
+ b4 * Traceability
$

Michiel

[Mariamishaq]

Thank you so much for your response.

Best,
Mariam Ishaq