ApolloChoiceModelling forum

Hi Stephane and David,

I have created my simple linear MNL model for the SP dataset and it works properly. However, when I have tried to convert the utility functions (cost and time) to log MNL, I am getting "NaN" values in my results for the following function.

apollo_probabilities(apollo_beta, apollo_inputs, functionality="estimate")

I have tried to remove the NA values from my dataset using the following code

rawData = read.csv("df_output_avAlts.csv",header=TRUE)
database <- rawData # Duplicate data frame
database[database == "" | database == " "] <- 0 # Replace blank & space by 0

But still, I am getting the following error for the model function:

--------------------------
Testing likelihood function...
Apollo found a model component of type MNL without a componentName. The name was set to "MNL" by
default.
Error in apollo_validate(mnl_settings, modelType, functionality, apollo_inputs) :
Some utilities for model component "MNL" contain values that are not finite numbers!
--------------------------

I was wondering if you could guide me on this.

Best wishes,
Nima

Nima

you're probably now taking the log of some values that are 0, so that would explain the failure

Stephane

Hi Stephane,

My utility function was including log for time and cost such as V = ASC + b_time * log(time) + b_cost * log(cost).

I have discussed it with David and I have changed the log(x) to log (1+x) for both time and cost as per his suggestion:
V = ASC + b_time * log(1 + time) + b_cost * log(1 + cost).

The problem is resolved now.

Thank you for your support.
Nima