Steps in generating efficient stated choice designs

This forum is for posts covering broader stated choice experimental design issues.

Moderators: Andrew Collins, Michiel Bliemer, johnr

Steps in generating efficient stated choice designs

Postby ericb » Fri Sep 20, 2019 12:21 pm

Hi there,

I am a relatively new user of nGene and I am a little bit confused about using nGene to produce efficient design.

I have used nGene previously to produce labelled and non labelled designs using priors derived from a mix of past research and gut feeling. I was fine with all this so far but recently came across Appendix 7B - Steps in generating efficient stated choice designs (page 131 in the manual) which list the following steps:

(1) specify the utility specifications for the likely final model
(2) randomly populate the design matrix to create an initial design
(3) calculate the choice probability of each alternative in the design (using draws because i have priors with distributions)
(4) constuct the AVC matrix etc...
(5) evaluate the statistical efficiency of the design
(6) go through a number of iterations (repeat 3 to 5)

Apologies if this is a stupid question but is the appendix describing what the software does or what I should be doing?! I am not sure whether I need to do anything with the information in this appendix.

Regards

E
ericb
 
Posts: 1
Joined: Thu Sep 19, 2019 2:45 pm

Re: Steps in generating efficient stated choice designs

Postby Michiel Bliemer » Fri Sep 20, 2019 1:27 pm

Ngene is used to generate an experimental design, all other steps are for the analyst to follow. As underlined in below, Ngene is used in steps 4 and 6.

I typically follow the following steps:

1. Formulate study objectives
Willingness-to-pay? Market shares? Elasticities?

2. Determine design dimensions
Which alternatives (labelled or unlabelled), which attributes, which attribute levels

3. Establish utility functions
Write them down on paper, including linear or nonlinear/qualitative (dummy/effects coding) variables and possible interactions.

4. Conduct a pilot study
Generate an initial experimental design using Ngene using the utility functions specified in step 3 assuming an orthogonal design or an efficient design with zero priors. Create a survey and get responses from some respondents.

5. Estimate initial parameters
Use the pilot data to estimate a simple multinomial logit model using estimation software such as Nlogit, Biogeme, or Apollo.

6. Generate an efficient design
Use the parameter estimates and their standard errors from step 5 as Bayesian priors and generate a Bayesian efficient design using Ngene.

7. Implement survey instrument
Use the experimental design created in step 6 to create choice tasks in a choice experiment, e.g. using online tools such as SurveyEngine, or Qualtrics, or Cofirmit other survey software.

Michiel
Michiel Bliemer
 
Posts: 1705
Joined: Tue Mar 31, 2009 4:13 pm


Return to Choice experiments - general

Who is online

Users browsing this forum: No registered users and 6 guests