LatentStage: Discrete heterogeneity model with multiple stages
In hollicam/DiscreteChoiceModels: Various models for marketing and social research

Description Usage Arguments Value Examples

The discrete heterogeneity model is primarily applicable to the scenario in which each subject is assigned a latent class, and fixed parameters are applied across those subjects in the same class. This particular function allows for more than one response, thus appropriate for constructing a model with multiple stages. Although each response is typically nested within one another, this property is not required, in order to fit the model.

1	LatentStage(nclass, ...)

`nclass`	Number of classes for all subjects, determining how many groups of different parameters will be obtained
`...`	X1, X2, X3... - Matrix of covariates for 1st, 2nd, 3rd stage, respectively. y1, y2, y3... - Vector of responses for 1st, 2nd, 3rd stage, respectively. id1, id2, id3... - Vector of subject IDs for 1st, 2nd, 3rd stage, respectively. The dimensions of these arguments must be the same, i.e. row(X1) = length(y1) = length(id1), as they determine the number of stages in the model.

A list object containing both arguments and results:

lambda - The estimate of class proportions, which sum up to 1.
beta - Estimates, SEs and p-values for all linear parameters.
posteriorz - List of probabilities of each subject belonging to a specific group, with class ID estimated by determining which one maximizes said probability.
all.loglik - List of log-likelihood estimates in each iteration.
y - List of responses for all stages.
id - List of subject IDs for all stages.
x - List of covariates for all stages.
AIC - Numerical AIC value for the current model.
BIC - Numerical BIC value for the current model.
runtime - Elapsed time in fitting the model.

data(threestage)
attach(threestage)
mod <- LatentStage(5, X1=stage1[, 4:7],
       y1=stage1$Y1, id1=stage1$Person,
       X2=stage2[, 4:7],
       y2=stage2$Y2, id2=stage2$Person,
       X3=stage3[, 4:7],
       y3=stage3$Y3, id3=stage3$Person)

data(dating)
attach(dating)
nonmiss <- !is.na(wrote)
mod <- LatentStage(3, y1 = browsed, y2 = wrote[nonmiss],
                   id1 = respid, id2 = respid[nonmiss],
                   X1 = agedif, X2 = agedif[nonmiss])