get_start: 'get_start' get starting values for discrete or continuous...
In nmm: Nonlinear Multivariate Models

Description Usage Arguments Value Examples

get_start get starting values for discrete or continuous choice model.

get_start(
  eq_c = NULL,
  eq_d = NULL,
  data = NULL,
  part = "joint",
  datan = "data",
  fixed_term = FALSE,
  weight_paths = TRUE,
  weight_paths_cont = FALSE,
  data_weight = NULL,
  par_c = NULL,
  par_d = NULL,
  best_method = FALSE,
  startvals = NULL,
  DEoptim_run = FALSE,
  hessian = NULL,
  transform = TRUE,
  MNtypef = "logit",
  pardogit = NULL,
  opt_method = "BFGS",
  numerical_deriv = FALSE
)

`eq_c`	Continuous equations errors.
`eq_d`	Discrete equations.
`data`	`data.frame` is used in the optimization.
`part`	Type of estimation: "joint", "cont", "disc".
`datan`	Name of `data.frame` used in the optimization.
`fixed_term`	if `TRUE`, includes fixed term in log-likelihood.
`weight_paths`	if `TRUE`, weights paths of the whole system.
`weight_paths_cont`	if `TRUE`, weight paths only in continuous part.
`data_weight`	`data.frame` with weights for continuous and discrete equations, same dim as `data`.
`par_c`	Names of parameters in continuous equations.
`par_d`	Names of parameters in discrete equations.
`best_method`	if `TRUE`, try all possible optimization methods and choose the one with the smallest likelihood.
`startvals`	Starting values, can be also `NULL`.
`DEoptim_run`	if `TRUE`, runs DEoptim for the optimization.
`hessian`	Name of hessian function.
`transform`	if `TRUE`, quantile transformation is applied.
`MNtypef`	"dogit" or "logit".
`pardogit`	"dogit" parameters.
`opt_method`	optimization method to use.
`numerical_deriv`	if `TRUE`, numerical derivatives are calculated in nmm function

Starting values for discrete or continuous blocks.

# Example of discrete choice model
data("TravelMode", package = "AER")
eq_d <- c("ASC1 * 1 + B2_t * travel_1 + B3_v * vcost_1" ,
          "ASC2 * 1  + B2_t * travel_2 + B3_v * vcost_2",
          "ASC3 * 1  + B2_t * travel_3 + B3_v * vcost_3",
          "ASC4 * 1  + B2_t * travel_4 + B3_v * vcost_4")
parl <- c(paste0("ASC", 1:4), "B2_t", "B3_v")
obj <- get_par(parl, eq_d)

mode_spec_var <- c("wait", "vcost", "travel", "gcost")
data <- TravelMode
data$wait <- as.numeric(data$wait)
data[data$wait==0,"wait"] <- 0.000001 # add a small number to 0
data$travel <- as.numeric(data$travel)
data[data$travel==0,"travel"] <- 0.000001
data$vcost <- as.numeric(data$vcost)
data[data$vcost==0,"vcost"] <- 0.000001
data <- prepare_data(data, choice="mode", dummy="choice", PeID="individual", WeID="",
type="long", mode_spec_var =mode_spec_var, wc=FALSE)
stv <- get_start(eq_d=eq_d, data=data, datan="data", part="disc", par_d = parl, 
transform = FALSE)

#example, system of equations
data("CreditCard", package="AER")
cdat <- CreditCard
cdat$income2 <- cdat$income^2
cdat$d_selfemp <- as.numeric(cdat$selfemp)
eq_c <- c("expenditure ~  b1*age + b2*income + b3*income2",
"income ~ a1*age + a2*d_selfemp + a3*dependents + a4*majorcards")
parl <- c(paste0("b", 1:3), paste0("a", 1:4))
para_cont <- get_par(parl, eq_c)
cheqs0 <- para_cont$cheqs0

stv <- get_start(eq_c = eq_c, data=cdat, datan="cdat", part="cont", par_c=parl)