EM: Estimate a mixture model using an EM algorithm

In rferrali/rogali: A series of functions for the corrupt_audit project

Description

Estimate a mixture model using an EM algorithm

Usage

EM(formula, data, penalization = 0, munCol = j, zCol = Z,
  yearCol = year, stateCol = state, tol = 0.001, mixTrace = 1,
  mixMaxit = 1000, pool = T, k = 10, nSamples = NULL,
  parallel = F, seed = NULL, cl = NULL, ...)

Arguments

formula	a formula expression as for regression models, of pooled form cbind(y1, y2, y3) ~ .
data	a data.frame in which to interpret the variables occurring in formula
penalization	a scalar between 0 and 1 to penalize observations whose type is unobserved. 1 discards those observations, 0 gives them weight equal to observations whose type is observed.
munCol	column in both data and dataMun containing the municipality id, syntax as in dplyr
zCol	column in both data and dataMun containing the (possibly missing) municipality type, syntax as in dplyr
yearCol	column in data containing the year, syntax as in dplyr
stateCol	column in data containing the state, syntax as in dplyr
tol	threshold used to determined convergence
mixTrace	report each mixTrace iteration Leave NULL for no reporting.
pool	should data be pooled automatically?
k	rounding parameter for numerical stability. Rounds to the order of log(-k)
nSamples	number of bootstrap replicates. If NULL, standard errors are not estimated.
parallel	should the bootstrap be parallelized? if TRUE, the cl argument shoud not be NULL
seed	optional seed
cl	cluster used for parallelization
...	additional parameters passed on to multinom
mixMaxIt	maximum number of iterations

Details

Make sure to check the methods man page to see related methods, as well as the posteriorprobs man page for how to deal with predicted probabilities.

rferrali/rogali documentation built on May 26, 2019, 7 p.m.