rkt {RegKink}

Estimation for regression kink with a time-varying threshold

Description

This is a function estimating regression kink with a time-varying threshold.

Usage

rkt(y,x,z,q,r01,r02,r11,r12,stp1,stp2)

Arguments

y

A vector of response.

x

A vector of regressor

z

A data matrix of control variables

q

A vector of variable affecting threshold

r01

Lower bounder of parameter space for r0

r02

Upper bounder of parameter space for r0

r11

Lower bounder of parameter space for r1

r12

Upper bounder of parameter space for r1

stp1

Step used in grid search of r0

stp2

Step used in grid search of r1

Value

A list with the elements

bols

The OLS estimates when a kink effect is ignored.

bt

The regression coefficients when a kink effect is included in the model.

gammahat0

The estimated threshold of the constant one in threshold parameters.

gammahat1

The estimated threshold of the slop in threshold parameters.

sig

The sum of squred errors of the kink model.

Examples

sta <- proc.time()
##Simulated data
set.seed(12345)
n=200
x = rnorm(n)
q = rnorm(n)
rt = 0.2 - 0.5*q
z = rnorm(n)
x1 = cbind(neg.part(x-rt),pos.part(x-rt),z)
b0 =c(1,2,1)
y = b0[1]*x1[,1]+b0[2]*x1[,2]+b0[3]*x1[,3]+ rnorm(n)



# set grid search paramaters
r01 = 0
r02 = 2
stp1 = 0.1
r11 = -10
r12 = 5
stp2 = 0.1
# estimate the model with a state-dependent threshold
est1 <- rkt(y,x,z,q,r01,r02,r11,r12,stp1,stp2)

proc.time() - sta