choice-metrics.com

Hi there,

I'm helping to design an unlabelled DCE related to job opportunities for post docs in the UK. As shown in the code below this message, we are using a d-efficient design.

There are seven attributes, two with four levels and five with three levels. We have included an opt out option. We have also included some small non-zero priors to indicate the ordering of the levels within each attribute.

When we've run the design for a little while and looked at the generated choice sets within the designs, we're still finding some dominant choice sets despite including an * for the different alternative in the ;alts line of code.

I was wondering if anyone had an explanation for this? It may well be that I've made a slight error with the code (despite consulting the manual!)

Thanks,

Tom

Code: Select all


design
;alts = opt1*, opt2*, OptOut
;block = 4
;rows = 48
;eff = (mnl, d)
;alg = mfederov
;con

;model:
U(opt1) =          b2_Contract.dummy[0.001|0.002|0.003]            * Contract[2,3,4,1]      ? Contract Length
                 + b3_Salary.dummy[0.001|0.002|0.003]             * Salary[2,3,4,1]        ? Salary
                 + b4_Location.dummy[-0.001|-0.002]                * Location[2,3,1]        ? Location
                 + b5_Interest.dummy[0.001|0.002]                  * Interest[2,3,1]        ? Research Interest
                 + b6_Development.dummy[0.001|0.002]               * Development[2,3,1]     ? Development Opportunities
                 + b7_Autonomy.dummy[0.001|0.002]                  * Autonomy[2,3,1]        ? Autonomy   
                 + b8_Culture.dummy[0.001|0.002]                   * Culture[2,3,1]         ? Research Culture       

/
                                                                       
U(opt2) =          b2_Contract.dummy * Contract
                 + b3_Salary.dummy * Salary
                 + b4_Location.dummy * Location
                 + b5_Interest.dummy * Interest
                 + b6_Development.dummy * Development
                 + b7_Autonomy.dummy * Autonomy
                 + b8_Culture.dummy * Culture

/

U(OptOut) = b1

$



I ran the script and checked in Excel and I could not detect any dominant alternatives when I consider the prior specification that you provided.

With your priors, contract length 4 is best and 1 is worst, salary 4 is best and 1 is worst, location 1 is best and 3 is worst, etc. So please check that you specified your priors correctly.
If you are convinced that they are correct, could you perhaps send me a choice task (row) of attribute levels that you believe has a dominant alternative?

Michiel

Hi Michiel,

Thanks as ever for your reply and for checking in Excel.

I will double check with my colleagues regarding the priors, as there has been some changing around of the levels earlier on in the project. This will probably be the issue as I can't see another issue with the code.

If this issue is still occurring I will get back to you.

Best wishes,

Tom