ApolloChoiceModelling forum

Dear Stephane,

just a general question to recheck:

If I have two different coefficients for two different attributes:

b_1 -0.54
b_2 -0.13

Can it be derived from that that the b_1 has more influence on the utility for the respondents so that it is more important to the respondents than b_2 (even if the attributes in the DCE have very different units; months and €)?

Thanks for your help.

hi

no, it cannot. All it tells you is that ONE UNIT of attribute 1 has more influence than ONE UNIT of attribute 2. So this all depends on the units, etc

Stephane

1. So it also cannot be said that they are more sensitive to the one attribute than to the other?

2. But within an Latent class model where b_1 is 0.69 for class A and b_1 is 0.12 for class B it can indeed be derived that for respondents in class A b_1 has higher positive influence on utility than for respondents of class B, right?

1. Again, it depends on the units. Imagine that I have b_cost = -0.1 and b_time = -0.05. You can't say that cost matters more than time as it depends on the units. Imagine that cost is in € and time is in minutes. Then what the above is telling us is that €1 is more important than 1 minute, but obviously not that cost is more important than time.

2. No, you cannot compare individual coefficients as there could be scale differences between classes. You can only compare relative values, so e.g. compare the relative impact of time over cost in class 1 with that in class 2

All this is covered in standard textbooks on choice modelling