ApolloChoiceModelling forum

Dear Everyone,

Hope this post finds you well. I have a query regarding producing predicted probability plots. Is it possible to produce a predicted probability plot for a predictor (keeping all other predictors to the mean) from an ICLV model in apollo? Is there any code example for producing predicted probability plots from an ICLV model? Thank you!

Hi

can you please be more specific with your query. Do you mean looking at the impact on probabilities of a change in one of the explanatory variables? If so, you can just use apollo_prediction after estimation and change the attribute in the database, and revalidate the inputs. It's not really different from what we do in Apollo_example_3.r

Stephane

Hi Dr. Hess,

Thank you for your kind reply. I appreciate it. Here is a simple example for demonstrating the issue.

V = list()
V[['no']] = 0
V[['yes']] = asc_yes + b_student*student+b_age* age +b_education*education_college+ b_income_L*income_low

### Define settings for MNL model component
mnl_settings = list(
alternatives = c(no=0, yes=1),
avail = 1,
choiceVar = Use,
V = V)

Here, if we have a binary choice of "yes" and "no" whether someone used a service or not. After estimating the coefficient of the model, if I get a coefficient of -0.24 for income_low indicating low income are less likely to use compared to not low-income individual. Now, I want to know how much less likely. So, I want to produce an effect plot of predicted probabilities of "yes" for both low-income and not low-income (keeping all the other predictors to their mean value). For logistics regression, there are options like "effects" package by Fox, et al. Is there any option I can do this type of plot with my ICLV model output from the apollo. Thank you!

Hi

yes, this is what you can do with apollo_prediction and changing the data. Please have a look at example_3 which does this for an MNL model and changes an attribute. The process is the same

Stephane

Hi Dr. Hess,

Thank you! I appreciate it.