choice-metrics.com

Dear Michiel,

I have an attribute, namely Parking Price. This attribute has fours levels: 0 (free of charge), 1$, 2$, no parking is available. I was wondering how to deal with the fourth level in the utility function in Ngene? I thought of adding a binary variable like Parking Availability and then put a constrain to ensure when there is no parking available the parking price is not applicable. The thing is that 0 is one of the attributes of parking price (free parking), so this binary variable may not work. Furthermore, there is a strong correlation between these two variables which is not recommended in choice set design. Would you please help me with this?

Thanks a lot for your massive support.
Best,
Peyman.

You could use an interaction effect:

U(altx) = ... + parking * PARKINGAVAILABLE[0,1] * PARKINGFEE[0,1,2] + ...

If PARKINGAVAILABLE = 0, then show "no parking available", if PARKINGAVAILABLE = 1, then show PARKINGFEE.

Michiel

Hi Michiel,

This is exactly the issue that I raised. When considering the interaction effect, in this case, there are two different situations that this term equals zero. The first one appears when no parking is available (B_parking * 0 (no parking) * PARKINGFEE[0,1,2] = 0) and the other one occurs when a parking is available but the parking price is 0 (B_parking * 1 (available parking) * 0 (free of charge) = 0). So, mathematically speaking, there is no difference between parking unavailability and a free of charge parking. While there is a significant difference in practice. In the first case, a person cannot park because there is no parking, but in the other case, he/she prefers to park as it is free. My question is how to make a distinction between these two occasions in Ngene?

Best,
Peyman.

Ah ok I thought you were asking how you specify an interaction effect in the Ngene syntax between parking fee and parking availability. So your question is not Ngene related, but rather related to how you formulate utility functions (which you also need for model estimation).

I believe this is what you want:

U(altx) = ... + parking * PARKINGAVAILABLE[0,1] + fee * PARKINGAVAILABLE * PARKINGFEE[0,1,2] + ...

Coefficient "parking" captures the parking availability, and coefficient "fee" captures the sensitivity to the parking fee.

Michiel