chngpt (version 2023.11-29)

sim.chngpt: Simulation Function

Description

Generate simulation datasets for change point Monte Carlo studies.

Usage

sim.chngpt (mean.model = c("thresholded", "thresholdedItxn",
 "quadratic", "quadratic2b", "cubic2b", "exp",
 "flatHyperbolic", "z2", "z2hinge", "z2segmented",
 "z2linear", "logistic"), threshold.type = c("NA",
 "M01", "M02", "M03", "M10", "M20", "M30", "M11",
 "M21", "M12", "M22", "M22c", "M31", "M13", "M33c",
 "hinge", "segmented", "upperhinge", "segmented2",
 "step", "stegmented"), b.transition = Inf, family =
 c("binomial", "gaussian"), x.distr = c("norm",
 "norm3", "norm6", "imb", "lin", "mix", "gam",
 "zbinary", "gam1", "gam2", "fixnorm", "unif"), e. =
 NULL, mu.x = 4.7, sd.x = NULL, sd = 0.3, mu.z = 0,
 alpha = NULL, alpha.candidate = NULL, coef.z =
 log(1.4), beta = NULL, beta.itxn = NULL,
 logistic.slope = 15, n, seed, weighted = FALSE,
 heteroscedastic = FALSE, ar = FALSE, verbose = FALSE)
sim.twophase.ran.inte(threshold.type, n, seed)
sim.threephase(n, seed, gamma = 1, e = 3, beta_e = 5, f = 7, beta_f = 2, coef.z = 1)

Value

A data frame with following columns:

y

0/1 outcome

x

observed covariate that we are interested in

x.star

unobserved covariate that underlies x

z

additional covariate

In addition, columns starting with 'w' are covariates that we also adjust in the model; columns starting with 'x' are covariates derived from x.

Arguments

threshold.type

string. Types of threshold effect to simulate, only applicable when label does not start with sigmoid.

family

string. Glm family.

n

n

mu.z

n

seed

seed

weighted

beta

beta

beta

coef.z

numeric. Coefficient for z.

beta.itxn

numeric. Coefficient for z.

alpha

numeric, intercept.

mu.x

numeric

sd.x

numeric

mean.model

numeric

x.distr

string. Possible values: norm (normal distribution), gam (gamma distribution). gam1 is a hack to allow e. be different

e.

e.

verbose

Boolean

b.transition

b.

sd

b.

ar

autocorrelation

alpha.candidate

Candidate values of alpha, used in code to determine alpha values

e

e

beta_e

beta_e

f

f

beta_f

beta_f

logistic.slope

beta_f

gamma

beta_f

heteroscedastic

Boolean.

Details

mean.model, threshold.type and b.transition all affect mean models.

Examples

seed=2
par(mfrow=c(2,2))
dat=sim.chngpt(mean.model="thresholded", threshold.type="hinge", family="gaussian", beta=0, n=200, 
    seed=seed, alpha=-1, x.distr="norm", e.=4, heteroscedastic=FALSE)
plot(y~z, dat)
dat=sim.chngpt(mean.model="thresholded", threshold.type="hinge", family="gaussian", beta=0, n=200, 
    seed=seed, alpha=-1, x.distr="norm", e.=4, heteroscedastic=TRUE)
plot(y~z, dat)
dat=sim.chngpt(mean.model="z2", threshold.type="hinge", family="gaussian", beta=1, n=200, 
    seed=seed, alpha=1, x.distr="norm", e.=4, heteroscedastic=FALSE)
plot(y~z, dat)
dat=sim.chngpt(mean.model="z2", threshold.type="hinge", family="gaussian", beta=1, n=200, 
    seed=seed, alpha=1, x.distr="norm", e.=4, heteroscedastic=TRUE)
plot(y~z, dat)