ApolloChoiceModelling forum

Hi, I am new in this forum, though not new as an apollo user (thanks by the way for this wonderful tool), but now I am struggling with a tricky task:

Is it possible with apollo to estimate a model, where each choice maker selects only one of 5 alternatives (i.e. single choice type), but we know from theory that these alternatives are related to each other in an ordinal manner? The ordered logit models on your example page seemed a promising option, but then I found out that this is a very different story, where each choice maker cunducts a full ranking of several alternatives. The model that I have in mind resembles more an ordinal regression.

Background: Based on a sample of non-owners and owners of private e-scooters, we would like to estimate a model according to the innovation decision process described by Rogers (2003), which proposes 5 ordinal stages of technology adoption. A tool often used in this context is poLCA, but I have limited trust in this tool, it seems like a black box that returns dramatically different results on minor changes in settings. Apollo is more robust, I think, and would also allow for latent classes.

Best wishes
Reinhard from BOKU

Reinhard

if a full ranking was given, you would use an exploded logit model. An ordered logit model is where the response to a single question is on an ordinal scale. If only first preference is given, it would be a multinomial choice. But maybe I don't understand your data

Stephane

Hi Stephane,

our respondents did not explicitly "choose" something. What actually happens ist that each respondent is assigned to a certain "stage of adoption" depending on several characteristics, including (i) ownership of an e-scooter (N/Y), (ii) for non-owners: level of knowledge and purchase intention, and (ii) for owners: frequency of usage. Now we want to investigate which more underlying personal characteristics and attitudes can explain why a person is on which stage (= rank). Conceptually, this corresponds to your second option (response to a single question on an ordinal scale).

Maybe you have already answered my question, I think I was confused by the terminology, I thought that the "ordered logit" is what you refer to as "exploded logit" for situations where the full rank is known. I have re-checked the guide and understand the explanation better know. Are you aware if there is an example for a simple ordered logit (not exploded) available on your example page?

Best
Reinhard

Reinhard

Yes, have a look at the OL.r example

Stephane