choice-metrics.com

Dear Prof.Michiel Blimer,
I have a limitation of sample size in my study around 200. The minimum of row is 6. My syntax is

Code: Select all

design
;alts=alt1*,alt2*,alt3
;rows=6
;eff=(mnl,d)
;con
;model:
U(alt1)=b0[0.000001]+b1.dummy[0.000002]*private[1,0]+b2.dummy[0.000002]*open[1,0]+b3[0.000002]*time[10,20,30]+b4.dummy[0.000002|0.000001]*channel[2,1,0]/
U(alt2)=b0[0.000001]+b1*private+b2*open+b3*time+b4*channel
$

Could I design row 12 with block 2 for 200 samples? Which one is better?

Sincerely your,
Sukunta

6 rows is really very little to estimate 6 coefficients, I would recommend to use a design of at least 12 rows. You can give all 12 choice tasks to a single respondent, or block it in 2 and give each block to 100 respondents. Using a larger design with more variation in the data is generally preferred over a very small design with limited variation.

Michiel

Dear Prof. Michiel Blimer,

Thank you so much for your recommendation. I have one question that the minimum of the row for design in equation S = k/n-1. Is it minimum for the choice set not suitable for the design?

Sincerely yours,
Sukunta

The minimum will work and you can estimate a model, but it may not be very efficient and it may be difficult to estimate more advanced models later on (such as mixed logit that have more parameters).

I refer to the work of Sandor and Wedel (2005), see https://journals.sagepub.com/doi/10.1509/jmkr.42.2.210.62285 and Kessels (2016), see https://www.sciencedirect.com/science/article/abs/pii/S1755534515300889.

Michiel

Dear Prof.Michiel Blimer,
Thank you so much.

Sincerely yours,
Sukunat