choice-metrics.com

Hi,

I am new user to Ngene, and I am still learning about the Stated Preference Surveys. I am aiming to conduct a stated preference survey related to transportation, asking the users about parking price against specific levels of attributes. As I am new to all of this, I appreciate any support regarding the following points:

1. I couldn't find the way of analyzing the data, after designing the survey and collecting the information from respondents. Please advise if Ngene has the capability to do so. Any example for such an exercise will help me a lot.

2. I am targeting to design my survey by using the efficient design, where I will assume the priors to generate my initial design then I will conduct a pilot survey. Please advise if I can use Ngene to analyse the outcome of the pilot survey to get a better estimation of the parameters and to revise and update the design accordingly. Any example for such an exercise will help me a lot.

3. It was stated in the manual that "Ngene allows the user to observe for these calculations (but only for panel MMNL designs) by clicking on the sample tick box within the RP panel tree structure", and the the sample generated was shown in "Figure 7.9: Panel MMNL simulated sample" from the manual. My question is how the Error value for each row from the generated sample as shown in Figure 7.9 is being calculated.


Thanks,
Bitoola.

Hi welcome to the forum.

1. Ngene is software to generate experimental designs for stated choice experiments. Ngene is not estimation software for finding parameters of choice models. For model estimation, you can use Nlogit (similar syntax to Ngene, commercial software), Biogeme (free software), R (free software), or other available software. For estimating simple multinomial logit models, you could possibly estimate parameters in Excel by specifying the loglikelihood function and use Solver.

2. See above, you will need to estimate parameters using pilot study data outside Ngene and feed the results back into Ngene.

3. The error is calculated by taking pseudo-random draws from the extreme value type I distribution, which is the distribution underlying logit models.

Michiel

Thanks for your feedback and prompt response.
Under the demonstration projects I found some CSV files called "Data Files". Would your please advise on the benefits of such feature?

Ngene can evaluate the efficiency of data e.g. the D-error of a given design or data. Data in Ngene consists of choice tasks with attribute levels (without choice observations).

Michiel

The information you have provided in this conversation and in the whole forum is so useful. I am designing a new survey related to parking, where participants is being asked about his/her preference between 2 options.

I have now a new issue, and I would appreciate your feedback if I want to add new alternative to the survey where the participant has the right not to select any option, like cancel my trip or change the destination. Should I include such option in the design noting that this alternative has no attributes or levels for the attributes? Will the utility function in this case be zero? Any advice how to include this in the analysis in R software?