choice-metrics.com

Hi Andrew, Michiel or Sir/Madam,

I was trying to create an orthogonal design using NGene, but got some outputs unusual. Below is a screenshot of the syntax and output by NGene.
As you may notice that I used orth=seq2 because each alternative has different attribute. The outputs for Auto and Bus are exactly the same (as highlighted). This can be avoided by using orth=sim, however.


Could you please provide any information that may cause this difference? Or the use of seq2 is inappropriate? Thanks in advance.

Chao

Syntax is attached here:

Design
;alts = auto, bus, sub, notrvl, chgdest
;rows = 9
;orth = seq2

;model:
U(auto) = ttauto * tt[0,1,2] + a_pcost * pcost[0,1,2] + var * var_auto[0,1,2] /
U(bus) = c_bus + ttbus * tt + b_fare * fare[0,1,2] + bfreq * serfreq[0,1,2] /
U(sub) = c_sub + ttsub * tt + s_fare * fare + sfreq * serfreq $

Dear Chao,

Since this is a labelled experiment with alternative specific coefficients, you want all attributes to be orthogonal (if orthogonality is of interest). In that case you need to use ;orth = sim.

I am not exactly sure why identical attribute levels are generated, we will look into it.