NGENE and Survey Engine

[Michiel Bliemer]

1. Ngene does not generate dominant alternatives when you specify the preference order of all attributes, together with the indication of unlabelled alternatives in the ;alts property. So there is no check needed. If you use zero priors you will manually need to check all choice tasks for any dominant alternatives (which do not occur if you have nominal attributes), simply by looking at them and assessing whether one attribute is always better than another attribute on every single attribute.

2. That is correct. You can add a constant during the model estimation phase to account for any possible left-to-right order bias when you present the choice tasks.

3. It is best to ask SurveyEngine, I can only answer questions about Ngene. Regarding simulation, I believe this simply checks whether your survey logic works, e.g. you can look at the data to make sure that no questions are skipped. Regarding utilities, they seem to use effects coding, so the sum of the utilities of each attribute is zero. The baseline in effects coding is the mean utility. You can easily convert to dummy coding yourself if you like. Larger utilities indicate more preference. The preference order in your screenshot seem the wrong way around, but this is perhaps based on simulated data, which is rubbish data and would not mean anything.

Michiel