LMest (version 3.2.2)

draw: Draw simulated sample from a Generalized Latent Markov Model

Description

Draw a sample for LMest objects of classes: LMbasic, LMbasiccont, LMlatent, LMlatentcont, and LMmixed

Usage

# S3 method for LMbasic
draw(est, n = NULL, TT = NULL, format = c("long","matrices"), seed = NULL, ...)
# S3 method for LMlatent
draw(est, n = NULL, TT = NULL, data, index, format = c("long", "matrices"),
                        fort = TRUE, seed = NULL, ...)
# S3 method for LMbasiccont
draw(est, n = NULL, TT = NULL, format = c("long","matrices"), seed = NULL, ...)
# S3 method for LMlatentcont
draw(est, n = NULL , TT = NULL, data, index, format = c("long", "matrices"),
                            fort = TRUE, seed = NULL, ...)
# S3 method for LMmixed
draw(est, n = NULL, TT = NULL, format = c("long", "matrices"), seed = NULL, ...)

Value

Y

matrix of response configurations unit by unit when est is of class LMbasic or LMmixed; array of continuous outcomes (n x TT x r) when est is of class LMbasiccont or LMlatentcont

S

matrix of distinct response configurations when est is of class LMbasic or LMmixed

yv

corresponding vector of frequencies when est is of class LMbasic or LMmixed

piv

vector of initial probabilities of the latent Markov chain when est is of class LMbasic

Pi

set of transition probabilities matrices (k x k x TT) when est is of class LMbasic

Psi

array of conditional response probabitlies (mb x k x r)when est is of class LMbasic

n

sample size

TT

number of time occasions

est

object of class LMbasic, LMlatent, LMbasiccont, LMlatentcont, or LMmixed

U

matrix containing the sequence of latent states (n x TT) when est is of class LMlatent or LMlatentcont

Psi

array of conditional response probabilities (mb x k x r) when est is of class LMlatent

Be

parameters affecting the logit for the initial probabilities when est is of class LMlatent or LMlatentcont

Ga

parametes affecting the logit for the transition probabilitieswhen est is of class LMlatent or LMlatentcont

latentFormula

a symbolic description of the model to be fitted when est is of class LMlatent. Detailed description is given in lmest

data

a data frame in long format, with rows corresponding to observations and columns corresponding to variables, a column corresponding to time occasions and a column containing the unit identifier when est is of class LMlatent or LMlatentcont

Mu

array of conditional means for the response variables (r x k) when est is of class LMlatentcont

Si

var-cov matrix common to all states (r x r) when est is of class LMlatentcont

latentFormula

a symbolic description of the model to be fitted. A detailed description is given in lmestCont

Arguments

est

object of class LMbasic (LMbasic-class), LMlatent (LMlatent-class), class LMbasiccont (LMbasiccont-class), LMlatentcont (LMlatentcont-class), or LMmixed (LMmixed-class)

n

sample size

format

character string indicating the format of final responses matrix

seed

an integer value with the random number generator state

data

a data frame in long format, with rows corresponding to observations and columns corresponding to covariates, a column corresponding to time occasions and a column containing the unit identifier when est is of class LMlatent or LMlatentcont

index

a character vector with two elements indicating the name of the "id" column as first element and the "time" column as second element when est is of class LMlatent or LMlatentcont

fort

to use fortran routine when possible (FALSE for not use fortran) when est is of class LMlatent or LMlatentcont

TT

number of time occasions when est is of class LMmixed

...

further arguments

Author

Francesco Bartolucci, Silvia Pandolfi, Fulvia Pennoni

Examples

if (FALSE) {
# draw a sample for 1000 units and only one response variable when est is of class LMbasic
n <- 1000
TT <- 6
k <- 2
r <- 1 #number of response variables
mb <- 3 #maximum number of response categories

piv <- c(0.7,0.3)
Pi <- matrix(c(0.9,0.1,0.1,0.9), k, k)
Pi <- array(Pi, c(k, k, TT))
Pi[,,1] <- 0
Psi <- matrix(c(0.7,0.2,0.1,0.5,0.4,0.1), mb, k)
Psi <- array(Psi, c(mb, k, r))
est = list(piv=piv, Pi=Pi, Psi=Psi, n=n, TT=TT)
class(est) = "LMbasic"

out <- draw(est)


data("data_SRHS_long")
SRHS <- data_SRHS_long[1:2400,]

SRHS$srhs <- 5 - SRHS$srhs

est <- lmest(responsesFormula = srhs ~ NULL,
             index = c("id","t"),
             data = SRHS,
             k = 3)

out1 <- draw(est = est, format = "matrices", seed = 4321, n = 100)

# draw a sample for 7074 units and only one response variable when est is of class LMlatent
data(data_SRHS_long)

data_SRHS_long$srhs <- 5 - data_SRHS_long$srhs
n <- length(unique(data_SRHS_long$id))
TT <- max(data_SRHS_long$t)

est  <- lmest(responsesFormula = srhs ~ NULL,
              latentFormula =  ~
              I(gender - 1) +
              I( 0 + (race == 2) + (race == 3)) +
              I(0 + (education == 4)) +
              I(0 + (education == 5)) +
              I(age - 50) + I((age-50)^2/100),
              index = c("id","t"),
              data = data_SRHS_long,
              k = 2,
              paramLatent = "multilogit",
              start = 0)

out <- draw(est = est, data = data_SRHS_long, index = c("id","t"),
            format = "matrices",seed = 4321)

est1 = list(Psi = est$Psi, Be = est$Be, Ga = est$Ga,
            paramLatent = "multilogit",n=n,TT=TT)

attributes(est1)$latentFormula = ~
              I(gender - 1) +
              I( 0 + (race == 2) + (race == 3)) +
              I(0 + (education == 4)) +
              I(0 + (education == 5)) +
              I(age - 50) + I((age-50)^2/100)
class(est1) = "LMlatent"

out1 <- draw(est = est1, data = data_SRHS_long, index = c("id","t"),
                     format = "matrices",
                     seed = 4321)

# draw a sample for 1000 units and 3 response variable when est is of class LMbasiccont
n <- 1000
TT <- 5
k <- 2
r <- 3 #number of response variables

piv <- c(0.7,0.3)
Pi <- matrix(c(0.9,0.1,0.1,0.9), k, k)
Pi <- array(Pi, c(k, k, TT))
Pi[,,1] <- 0
Mu <- matrix(c(-2,-2,0,0,2,2), r, k)
Si <- diag(r)
est = list(piv=piv,Pi=Pi,Mu=Mu,Si=Si,n=n,TT=TT)
class(est) = "LMbasiccont"

out <- draw(est)

data(data_long_cont)

est <- lmestCont(responsesFormula = Y1 + Y2 + Y3 ~ NULL,
                 index = c("id", "time"),
                 data = data_long_cont,
                 k = 3,
                 modBasic = 1,
                 tol = 10^-5)

out2 <- draw(est = est, n = 100, format = "long", seed = 4321)

# draw a sample for 1000 units and 3 response variable when est is of class LMlatentcont
data(data_long_cont)

est <- lmestCont(responsesFormula = Y1 + Y2 + Y3~ NULL,
                 latentFormula = ~ X1 + X2,
                 index = c("id", "time"),
                 data = data_long_cont,
                 k = 3)
out <- draw(est = est, data = data_long_cont, index = c("id", "time"), format = "matrices", 
            seed = 4321)

est1 <- list(Mu = est$Mu,Si = est$Si,Be = est$Be,Ga = est$Ga,paramLatent="multilogit",n=est$n,
             TT=est$TT)
attributes(est1)$latentFormula = ~ X1 + X2
class(est1) = "LMlatentcont"
out1 <- draw(est = est1, data = data_long_cont,
                         index = c("id", "time"),
                         fort=TRUE, seed = 4321, format = "matrices")
}

# draw a sample for 1000 units and only one response variable and 5 time occasions 
# when est if of class LMmixed
k1 <- 2
k2 <- 3
la <- rep(1/k1, k1)
Piv <- matrix(1/k2, k2, k1)
Pi <- array(0, c(k2, k2, k1))
Pi[,,1] <- diag(k2)
Pi[,,2] <- 1/k2
Psi <- cbind(c(0.6,0.3,0.1), c(0.1,0.3,0.6), c(0.3,0.6,0.1))
est <- list(la=la, Piv=Piv, Pi=Pi, Psi=Psi, n=1000,TT=5)
class(est) = "LMmixed"

out <- draw(est = est)

if (FALSE) {
# Example based on criminal data when est if of class LMmixed
data(data_criminal_sim)
data_criminal_sim = data.frame(data_criminal_sim)

# Estimate mixed LM model for females
responsesFormula <- lmestFormula(data = data_criminal_sim,
                                  response = "y")$responsesFormula
est <- lmestMixed(responsesFormula = responsesFormula,
                  index = c("id","time"),
                  k1 = 2,
                  k2 = 2,
                  data = data_criminal_sim[data_criminal_sim$sex == 2,])

out <- draw(est = est, n = 100, TT = 6, seed = 4321)
}