To do that, I wanted to develop a latent class, where the classes differ in the number of alternatives that are considered.
I have modified the LC model in order to do that.
I first, create a list of list for the describing the availability of the different alternatives for the different classes.
Then I tried to modify the code inside the probability function, so that it takes the availability list specific for each class.
However, doing this I hit an error:
Error in apollo_mnl(mnl_settings, functionality) :
Duplicated componentName found (EClass_1). Names must be different for each component.
I do not see what error I made, and whether what I am doing is sound?
I can also share a small dataset to run the model. But I am not sure that I can attach a file to this message.
Damien
Here is the full code:
Code: Select all
rm(list = ls())
apollo_initialise()
# description of the model
apollo_control = list(
modelName = "EBA_subsidy",
indivID = "interview__key",
outputDirectory = "output"
)
load(file = "forForum.RData")
# Vector of parameters, including any that are kept fixed in estimation
# Hypothesis 3 classes (1 RUM 1 EBA Work 1 EBA fodd)
# Preferences are different for each class...
apollo_beta=c(b_ASC = -2.3,
b_work_a = -0.5,
b_work_b = 0 ,
b_work_c = +0.5 ,
b_fodd_a = 0.007,
b_fodd_b = 0.005,
b_fodd_c = 0.1,
b_legu_a = 0.1,
b_legu_b = 0.1,
b_legu_c = 0.1,
b_engr_a = -0.1,
b_engr_b = -0.1,
b_engr_c = -0.1,
b_subs_a = -0.1,
b_subs_b = -0.1,
b_subs_c = -0.1,
delta_b = 0,
delta_c = 0)
# Vector with names (in quotes) of parameters to be kept fixed at their starting value in apollo_beta, use apollo_beta_fixed = c() if none
apollo_fixed = c( )
# ################################################################# #
#### DEFINE LATENT CLASS COMPONENTS ####
# ################################################################# #
apollo_lcPars=function(apollo_beta, apollo_inputs){
lcpars = list()
lcpars[["beta_work"]] = list(b_work_a, b_work_b, b_work_c)
lcpars[["beta_fodd"]] = list(b_fodd_a, b_fodd_b, b_fodd_c)
lcpars[["beta_legu"]] = list(b_legu_a, b_legu_b, b_legu_c)
lcpars[["beta_engr"]] = list(b_engr_a, b_engr_b, b_engr_c)
lcpars[["beta_subs"]] = list(b_subs_a, b_subs_b, b_subs_c)
### Utilities of class allocation model
V=list()
V[["class_a"]] = 0
V[["class_b"]] = delta_b
V[["class_c"]] = delta_c
### Settings for class allocation models
classAlloc_settings = list(
classes = c(class_a=1, class_b=2, class_c=3),
avail = list(class_a=1, class_b=1, class_c=1),
utilities = V,
componentName = "AllocClass"
)
lcpars[["pi_values"]] = apollo_classAlloc(classAlloc_settings)
return(lcpars)
}
# Preparing lists of availability for each class of EBA
availa = list()
availa[["EClass_1"]] = list("1"=1, "2"=1, "SQ"=1)
availa[["EClass_2"]] = list("1"=database$avail1Work, "2"=database$avail2Work, "SQ"=1) ## EBA Work
availa[["EClass_3"]] = list("1"=database$avail1Fodd, "2"=database$avail2Work, "SQ"=1)
# Checking that all that is needed for the model is here i.e. apollo_control, database, apollo_beta & apollo fixed
apollo_inputs = apollo_validateInputs()
# apollo_probabilities
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()
### Loop over classes
for(s in 1:3){
mnl_settings = list(
alternatives = c("1"=1, "2"=2, "SQ"=3),
avail = availa[[s]],
choiceVar = CHOICE,
componentName = paste0("EClass_",s)
)
print(mnl_settings$componentName)
### Compute class-specific utilities
V=list()
V[["SQ"]] = b_ASC + beta_work[[s]]*WORKSQ/100 + beta_fodd[[s]]*FODDSQ/10 + beta_legu[[s]]*LEGUSQ + beta_engr[[s]]*ENGRSQ/10 + beta_subs[[s]]*SUBSSQ/10
V[["1"]] = beta_work[[s]]*WORK1/100 + beta_fodd[[s]]*FODD1/10 + beta_legu[[s]]*LEGU1 + beta_engr[[s]]*ENGR1/10 + beta_subs[[s]]*SUBS1/10
V[["2"]] = beta_work[[s]]*WORK2/100 + beta_fodd[[s]]*FODD2/10 + beta_legu[[s]]*LEGU2 + beta_engr[[s]]*ENGR2/10 + beta_subs[[s]]*SUBS2/10
mnl_settings$utilities = V
### Compute within-class choice probabilities using MNL model
P[[paste0("EClass_",s)]] = apollo_mnl(mnl_settings, functionality) + 0.001
### Take product across observation for same individual
P[[paste0("EClass_",s)]] = apollo_panelProd(P[[paste0("EClass_",s)]], apollo_inputs ,functionality)
}
### Compute latent class model probabilities
lc_settings = list(inClassProb = P, classProb=pi_values)
P[["model"]] = apollo_lc(lc_settings, apollo_inputs, functionality)
### Prepare and return outputs of function
P = apollo_prepareProb(P, apollo_inputs, functionality)
return(P)
}
model = apollo_estimate(apollo_beta, apollo_fixed,
apollo_probabilities, apollo_inputs)
apollo_modelOutput(model)