ApolloChoiceModelling forum

Hi,
I am trying to run hybrid choice model for a choice experiment data. I followed your example code and manual to set up my model. Unfortunately, I keep getting error message "Log-likelihood calculation fails at values close to the starting values!
In addition: Warning message:
In log(P[[j]]) : NaNs produced"

When I tried "apollo_beta=apollo_searchStart(apollo_beta, apollo_fixed,apollo_probabilities, apollo_inputs)" to find initial values of beta, I get the following error message:
"Testing probability function (apollo_probabilities)
Error: node stack overflow
Execution halted"

Could you please give me suggestions what I can do to address these? If required, I can share some description of the data and code. Thank you.

Hi,

Could you please share your code? It is hard to diagnose without looking at the actual code.

Cheers
David

Hi,

Thank you very much for the reply. Please see my code below:


### Description of variables
#Choice variable: binary: Status quo (SQ or S), Alternatives (Alt or A)
#Attributes: WeaAbsLin SpeAbsLin NatAbsLin Levy
#Explanatory variables: HHsize, IncCat_Mid, IncCat_High
#Indicator variable: attitude_Epl (rank 1 to 7)


### Clear memory


rm(list = ls())

#### LOAD DATA AND APPLY ANY TRANSFORMATIONS ####
# ################################################################# #



### Load Apollo library
#install.packages("apollo", repos = "http://cran.us.r-project.org")
library(apollo)

### Initialise code
apollo_initialise()

### Set core controls
apollo_control = list(
modelName = "hybrid_model_WTP",
modelDescr = "Hybrid choice model on LCF_WTP",
indivID = "id",
mixing = TRUE,
nCores = 25
)

load("database.RData")

# ################################################################# #



# ################################################################# #
#### ANALYSIS OF CHOICES ####
# ################################################################# #


choiceAnalysis_settings <- list(
alternatives = c(SQ="S",Alt="A"),
avail = list(SQ=1, Alt=1),
choiceVar = database$Choice,
explanators = database[,c("HHsize", "IncCat_Mid", "IncCat_High" )]
)
apollo_choiceAnalysis(choiceAnalysis_settings, apollo_control, database)



# ################################################################# #
#### DEFINE MODEL PARAMETERS ####
# ################################################################# #

### Vector of parameters, including any that are kept fixed in estimation
### Vector of parameters, including any that are kept fixed in estimation
apollo_beta <- c(b_Wea = 0,
b_Spe = 0,
b_Nat = 0,
b_Levy = 0,
b_IncMid = 0,
b_IncHigh = 0,
gamma_Wea_HHsize = 0,gamma_Spe_HHsize = 0,
gamma_Nat_HHsize = 0,
gamma_Wea_MidI = 0,gamma_Spe_MidI = 0,
gamma_Nat_MidI = 0,
gamma_Wea_HighI = 0,gamma_Spe_HighI = 0,
gamma_Nat_HighI = 0,
gamma_LV_HHsize = 0, gamma_LV_MidI = 0,
gamma_LV_HighI = 0,
mu_log_asc = 0,
sigma_log_asc = 1,lambda= 1,
zeta_Epl=1,
tau_Epl_1 =-3, tau_Epl_2 =-3,
tau_Epl_3 = -1, tau_Epl_4 = 1,tau_Epl_5 = 2,
tau_Epl_6 = 3
)
apollo_fixed <- c("mu_log_asc", "sigma_log_asc")


# ################################################################# #
#### DEFINE RANDOM COMPONENTS ####
# ################################################################# #

### Set parameters for generating draws
apollo_draws = list(
interDrawsType="halton",
interNDraws=500,
interNormDraws=c("draws_asc_inter","eta")
)

### Create random parameters
apollo_randCoeff <- function(apollo_beta, apollo_inputs) {
randcoef <- list()
randcoef[['asc']] <- mu_log_asc + b_IncMid * (IncCat=='Mid') + b_IncHigh * (IncCat=='High')-sigma_log_asc*draws_asc_inter
randcoef[["LV"]] = gamma_LV_HHsize*HHsize + gamma_LV_MidI*IncCat_Mid+ gamma_LV_HighI*IncCat_High + eta

return(randcoef)
}



# ################################################################# #
#### GROUP AND VALIDATE INPUTS ####
# ################################################################# #

apollo_inputs = apollo_validateInputs()

# ################################################################# #
#### DEFINE MODEL AND LIKELIHOOD FUNCTION ####
# ################################################################# #
#options(expressions=500000) ## avoid "node stack overflow" error
apollo_probabilities=function(apollo_beta, apollo_inputs, functionality="estimate"){

### Attach inputs and detach after function exit
apollo_attach(apollo_beta, apollo_inputs)
on.exit(apollo_detach(apollo_beta, apollo_inputs))

### Create list of probabilities P
P = list()

### Create alternative specific constants
asc_SQ_value = 0
asc_Alt_value = asc

### Likelihood of choices
### List of utilities: these must use the same names as in mnl_settings, order is irrelevant
V <- list()
V[['SQ']] <- asc_SQ_value +
b_Wea * WeaAbsLin_S + b_Spe * SpeAbsLin_S + b_Nat * NatAbsLin_S+ lambda*LV
V[['Alt']] <- asc_Alt_value +
b_Wea * WeaAbsLin_A + b_Spe * SpeAbsLin_A + b_Nat* NatAbsLin_A + b_Levy * Lev_A + lambda*LV


### Define settings for MNL model component

mnl_settings = list(
alternatives = c(SQ="S",Alt="A"),
avail = list(SQ=1, Alt=1),
choiceVar = Choice,
V = V
)


### initial values of beta: when I use this I find error message "node stack overflow"

#apollo_beta=apollo_searchStart(apollo_beta, apollo_fixed,apollo_probabilities, apollo_inputs)
#apollo_beta=apollo_bootstrap(apollo_beta, apollo_fixed, apollo_probabilities, apollo_inputs)


### Compute probabilities for MNL model component
P[["Choice"]] = apollo_mnl(mnl_settings, functionality)

### Likelihood of indicators
ol_settings1 = list(outcomeOrdered=attitude_Epl,
V=zeta_Epl*LV,
tau=c(tau_Epl_1, tau_Epl_2, tau_Epl_3, tau_Epl_4,tau_Epl_5,tau_Epl_6),
rows=(task==1))


P[["indic_Epl"]] = apollo_ol(ol_settings1, functionality)



### Likelihood of the whole model
P = apollo_combineModels(P, apollo_inputs, functionality)

### Take product across observation for same individual
P = apollo_panelProd(P, apollo_inputs, functionality)

### Average across inter-individual draws
P = apollo_avgInterDraws(P, apollo_inputs, functionality)

### Prepare and return outputs of function
P = apollo_prepareProb(P, apollo_inputs, functionality)
return(P)
}

# ################################################################# #
#### CHECK FOR COMPUTATIONAL REQUIREMENTS ####
# ################################################################# #

speedTest_settings=list(
nDrawsTry = c(250, 500, 1000),
nCoresTry = 1:3,
nRep = 10
)

apollo_speedTest(apollo_beta, apollo_fixed, apollo_probabilities, apollo_inputs, speedTest_settings)

# ################################################################# #
#### MODEL ESTIMATION ####
# ################################################################# #

### Estimate model
model = apollo_estimate(apollo_beta, apollo_fixed, apollo_probabilities, apollo_inputs)

# ################################################################# #
#### MODEL OUTPUTS ####
# ################################################################# #

# ----------------------------------------------------------------- #
#---- FORMATTED OUTPUT (TO SCREEN) ----
# ----------------------------------------------------------------- #

apollo_modelOutput(model)

# ----------------------------------------------------------------- #
#---- FORMATTED OUTPUT (TO FILE, using model name) ----
# ----------------------------------------------------------------- #

apollo_saveOutput(model)

Hi,

It seems the starting value of the OL thresholds are invalid. You have:
tau_Epl_1 =-3, tau_Epl_2 =-3,
tau_Epl_3 = -1, tau_Epl_4 = 1,tau_Epl_5 = 2,
tau_Epl_6 = 3

But they need to be all different and increasing, so if you change it to:
tau_Epl_1 =-3, tau_Epl_2 =-2,
tau_Epl_3 = -1, tau_Epl_4 = 1,tau_Epl_5 = 2,
tau_Epl_6 = 3

It should work.

Also, if you define the thresholds as a list instead of a vector, the model might run faster, as follows.
ol_settings1 = list(outcomeOrdered=attitude_Epl,
V=zeta_Epl*LV,
tau=list(tau_Epl_1, tau_Epl_2, tau_Epl_3, tau_Epl_4,tau_Epl_5,tau_Epl_6),
rows=(task==1))

Best
David

Hi David,

It works. Thank you very much for your help.

Zabid