zi_fit_R: Fits the Hurdle model assuming linear abk parametrization...

In sqyu/ZiDAG: Directed Graphical Models and Causal Discovery for Zero-Inflated Data

Description

Fits the Hurdle model assuming linear abk parametrization using stats::optim.

Usage

zi_fit_R(
  V,
  Y,
  left,
  right,
  maxit = 200,
  tol = 1e-08,
  runs = 1,
  value_only = TRUE,
  report = 10
)

Arguments

V	A matrix of 0/1s, equal to Y != 0.
Y	A data matrix of the same size as V.
left	An integer between 1 and ncol(Y). The index of the variable to be fit.
right	A vector of integers between 1 and ncol(Y) different from left. Indices of the "regressors".
maxit	An integer, the maximum number of integers, argument to stats::optim.
tol	A positive number, the tolerance passed as the pgtol argument to stats::optim.
runs	A positive integer, number of reruns; if larger than 1, stats::optim will be called multiple times with different initial values, and the fit that maximizes the log likelihood will be used.
value_only	If TRUE, returns the minimized negative log likelihood only. Defaults to TRUE.
report	An integer indicating verbosity, argument to stats::optim. If set to 1 no output is print during estimation.

Details

Fits the Hurdle model assuming linear abk parametrization, where Y[,left] conditional on Y[,right] is a 1-d Hurdle model with respect to the sum of the Lebesgue measure and a point mass at 0 with density a*v+b*y-y^2*k/2-log(1+sqrt(2pi/k)*exp(a+b^2/(2k))), with a and b both linear functions in V[,right] and Y[,right].

Value

If value_only == TRUE, returns the minimized negative log likelihood only. Otherwise, returns

nll	A number, the minimized negative log likelihood.
par	A vector of length 4*length(right)+3, the fitted parameters, in the other of: the intercept for the a (a scalar), linear coefficients on V[,right] for a, linear coefficients on Y[,right] for a, the intercept for the b (a scalar), linear coefficients on V[,right] for b, linear coefficients on Y[,right] for b.
n	An integer, the sample size.
effective_df	4*length(right)+3, the effective degree of freedom.

Examples

m <- 3; n <- 1000
adj_mat <- make_dag(m, "complete")
dat <- gen_zero_dat(1, "abk", adj_mat, n, k_mode=1, min_num=10, gen_uniform_degree=1)
zi_fit_R(dat$V, dat$Y, 3, 1:2, maxit=1000, runs=2, value_only=TRUE, report=0)
zi_fit_R(dat$V, dat$Y, 3, 1:2, maxit=1000, runs=2, value_only=FALSE, report=0)

sqyu/ZiDAG documentation built on Jan. 19, 2021, 4:11 p.m.