choice-metrics.com

1) Thank you very much Michiel! This is very helpful. I work on revising my ATTR. levels to allow for some flexibility.

- Indeed, I forgot to remove the constraints of type altx.mode = altx.mode that were there from my previous syntax, but are not necessary here.

- Yes, by ;reject altx.mode = 2 and altx.cost <> 0 I mean that altx.cost is 0 for mode 2 (== bike). There are many of such constraints as I need to define each of them for all 5 alternatives and all 12 modes.

2) The constraints in my second question are a bit complicated. As the levels of the cost attribute are calculated for indicative distances in the case of car (mode == 1), taxi (mode == 3) and ride-sharing (mode == 4), I want to have one to one correspondence for the costs of these 3 modes; meaning that when the cost for car is 0.5, the costs for taxi and ride-sharing should be 1.5 and 1.0, respectively. But again this is quite restrictive, so I guess the solution is to allow for various cost levels per distance.

Best,
Eva

If the constraints become too complicated for Ngene to automatically account for all of them, you can of course also consider making some manual changing afterwards. Making small changes by hand is fine as long as you do not create perfect correlations between attributes. After making manual changes you can always evaluate the design in Ngene and see if there are any issues (if there are issues, then the D-error will be infinite, while if the D-error is a finite number then the model can be estimated).

Michiel

I will try that as well, as after increasing the range of the attribute levels and relaxing some constrains I still cannot get a solution. Thank you!