hinge.test: A non-nested hypothesis testing problem for threshold...

In chngpt: Estimation and Hypothesis Testing for Threshold Regression

A non-nested hypothesis testing problem for threshold regression models

Description

Test a hinge effect against a linear effect

Usage

hinge.test(formula, cov.interest, family = c("binomial", "gaussian"), data, thres = NA,
    lb.quantile = 0.1, ub.quantile = 0.9, chngpts.cnt = 10, method = c("FDB", "B", "DB"),
    boot.B = 10000, B2 = NA, verbose = FALSE)

Arguments

formula
cov.interest
family
data
thres	If supplied, this will be the threshold value to use in the hinge model.
lb.quantile	lower bound of threshold candidates in quantile
ub.quantile	upper bound of threshold candidates in quantile
chngpts.cnt	number of candidate thresholds
method	type of test. FDB: false double bootstrap, B: parametric bootstrap, DB: double bootstrap.
boot.B	number of parametric bootstrap replicates for B and FDB
B2	number of inner bootstrap replicates for DB
verbose

Value

A list of class htest

p.value	P-value
chngpts	Vector of change points evaluated
TT	Standardized absolute score statistics
V.S.hat	Estimated variance-covariance matrix of the score statistics

Author(s)

Zonglin He

References

He, Fong, Fouda, Permar. A non-nested hypothesis testing problem for threshold regression model, under review

Examples

dat=sim.hinge(threshold.type = 'NA',family = 'binomial',thres='NA',X.ditr = 'norm',mu.X = c(0,0,0),
    coef.X = c(0,.5,.5,.4),cov.X = diag(3),eps.sd = 1,seed = 1,n=100)
test=hinge.test(Y~X1+X2, "x", family="binomial", data=dat,'method'='FDB',boot.B=10)
test

chngpt documentation built on Feb. 16, 2023, 10:38 p.m.