Type: Package
Title: Threshold Model and Unit Root Tests in Cross-Section and Time Series Data
Version: 1.9.3
Date: 2024-08-18
Author: Ho Tsung-wu [aut, cre]
Maintainer: Ho Tsung-wu <tsungwu@ntnu.edu.tw>
Description: Threshold model, panel version of Hylleberg et al. (1990) <doi:10.1016/0304-4076(90)90080-D> seasonal unit root tests, and panel unit root test of Chang (2002) <doi:10.1016/S0304-4076(02)00095-7>.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]
LazyData: TRUE
LazyLoad: yes
Depends: R (≥ 3.5.0)
Imports: plm, pglm, lme4, glmmTMB
NeedsCompilation: no
Packaged: 2024-08-18 12:44:32 UTC; badal
Repository: CRAN
Date/Publication: 2024-08-18 13:10:01 UTC

Panel Data Regression: Threshold Model and Unit Root Tests

Description

Functions for analysis of panel data, including the panel threshold model of Hansen (1999,JE), panel unit root test of Chang(2002,JE) based upon instuments generating functions (IGF), and panel seasonal unit root test based upon Hylleberg et al.(1990,JE).

Details

This version offers formatted output. This package designs a specification function ptm() to estimate the panel threshold model of Hansen(1999). The key feature of ptm() is to generalize Hansen's original code to allow multiple (more-than-one) regime-dependent right-hand-side independent variables; Dr. Hansen's original code admits only 1 regime-dependent right-hand-side independent variable. This version also includes panel unit root tests based on the instrument generating functions(IGF), proposed by Chang (2002, J. of Econometrics), and the panel version of Hylleberg et al.(1990) seasonal unit root test, proposed by Otero, et al. (2005, 2007).

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, College of Management, National Taiwan Normal University.
Maintainer: Ho Tsung-wu <tsungwu@ntnu.edu.tw>

References

Chang, Y. (2002) Nonlinear IV Unit Root Tests in Panels with Cross-Sectional Dependency. Journal of Econometrics, 110, 261-292.
Hansen B. E. (1999) Threshold effects in non-dynamic panels: Estimation, testing and inference. Journal of Econometrics,93, 345-368.
Hylleberg, S., Engle, R.F., Granger, C.W.J., and Yoo, B.S.(1990) Seasonal integration and cointegration. Journal of Econometrics,44, 215-238.
Otero, J., Smith, J., and Giulietti, M. (2005) Testing for seasonal unit roots in heterogeneous panels. Economics Letters, 86, 229-235.
Otero, J., Smith, J., and Giulietti, M. (2007) Testing for seasonal unit roots in heterogeneous panels in the presence of cross section dependence. Economics Letters, 86, 179-184.
Pesaran M. Hashem (2007) A simple panel unit root test in the presence of cross-section dependence. Journal of Applied Econometrics, 22, 265-312.

Seasonal unit root test based on Hylleberg et al. (1990)

Description

The function performs seasonal unit root test based on Eq.(3.6) of Hylleberg et al. (1990), univariate time series.

Usage

HEGY.test(wts, itsd, regvar = 0, selectlags = list(mode = "signf", Pmax = NULL))

Arguments

wts

Univariate time series

itsd

Options for c(i,t,sd)
i=1, intercept;=0 no intercept
t=1, trend;=0 no deterministic trend
sd=1, season dummy 1:(s-1);=0 no.

regvar

Additional regressors

selectlags

Selection of lags
mode, Criteria for selection, having three options: "signf","bic","aic".
Pmax, maximum number of lags.

Details

Mode for selectlags has three options, AIC and BIC use R built-in functions for linear model and their meanings are popular and straightforward. They include only lags that meet specific criterion, others are dropped from regressors. That is, lag orders of your model may not be a regular sequence. See also selPsignf() and selPabic().

Value

stats

Tests statistics for HEGY regression coefficients.

hegycoefs

HEGY regression coefficients.

lagsorder

Lags order. "aic" or "bic" returns a scalar; "signf" returns a sequence of numbers

lagcoefs

Coefficients of lag terms.

regvarcoefs

Coefficient(s) of additional regressor(s).

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle

References

Hylleberg, S., Engle, R.F., Granger, C.W.J., and Yoo, B.S.(1990) Seasonal integration and cointegration. Journal of Econometrics,44, 215-238.
Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

Examples

data(inf_Q)
y<-inf_Q[,1]
hegy.out<-HEGY.test(wts=y, itsd=c(1,0,c(1:3)),regvar=0, selectlags=list(mode="aic", Pmax=12))

hegy.out$stats #HEGY test statistics
names(hegy.out) # HEGY objects, which can be called by using $, see below.
hegy.out$hegycoefs
hegy.out$regvarcoefs

Unit root test based on Change(2002)

Description

This function estimates the unit root regression based on instrument generating function of Change(2002) and returns useful outputs.

Usage

IGF(y, maxp, ic, spec)

Arguments

y

A univariate time series data

maxp

the max number of lags

ic

Information criteria, either "AIC" or "BIC"

spec

regression model specification.
=0, no intercept and trend.
=1, intercept only.
=2, intercept and trend.

Details

Estimate univariate unit root test of Chang(2002).

Value

tstat.IGF

IGF unit root test

beta

regression coefficients. The first one is the AR(1) coefficient of unit root, and the last one is the intercept or trend

sdev

The IGF standard error for unit root coefficient

cV

The scalar C in IGF equation

p

The optimal number of lag

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, College of Management, National Taiwan Normal University.

References

Chang, Y. (2002) Nonlinear IV Unit Root Tests in Panels with Cross-Sectional Dependency. Journal of Econometrics, 110, 261-292.

Examples

data(inf19)
y <- inf19[,1]
IGF(y,maxp=35,ic="BIC",spec=2)$tstat.IGF

Estimation of sub-sampled data

Description

A function for estimating the subsampled data.

Usage

SMPLSplit_est(data,dep,indep,th,plot,h=1,nonpar=2)

Arguments

data

the data in either data.frame or matrix.

dep

the name of dependent variable.

indep

the name(s) of independent variable(s).

th

the name of threshold variable.

plot

=1, plot; =0, do not plot.

h

h=1, heteroskedasticity-consistent covariance; h=0, homoskedastic case.

nonpar

Indicator for non-parametric method used to estimate nuisance scale in the presence of heteroskedasticity (only relevant if h=1).Set nonpar=1 to estimate regressions using a quadratic.Set nonpar=2 to estimate regressions using an Epanechnikov kernel with automatic bandwidth.

Details

This code estimates the parameters of sub-sampled data. It generalizes the simple code of Dr. Hansen, allowing White Corrected Heteroskedastic Errors.

Value

threshold

values of threshold estimates.

est0

coefficient estimates of global data.

est.low

coefficient estimates of low regime.

est.high

coefficient estimates of high regime.

est0.info

additional information of global data.

est.joint.info

additional information of joint threshods.

est.low.info

additional information of est.low.

est.high.info

additional information of est.high.

Note

Original code offered by Dr. B. E.Hansen (http://www.ssc.wisc.edu/~bhansen/).

References

Hanse B. E. (2000) Sample Splitting and Threshold Estimation. Econometrica, 68, 575-603.

Examples

## Not run, becasue of bootstrap replicaiton takes time. Users may unmark # and run. 
data("dur_john")
rep <- 500
trim_per <- 0.15
dep <- "gdpGrowth"
indep <- colnames(dur_john)[c(2,3,4,5)]

SMPLSplit_est(data=dur_john,dep,indep,th="GDP60",plot=0,h=1,nonpar=2)

Example code for sample splitting

Description

A sample code for learning sample splitting.

Usage

SMPLSplit_example(data,dep,indep,th1,th2,trim_per,rep,plot)

Arguments

data

the data in either data.frame or matrix.

dep

the name of dependent variable.

indep

the name(s) of independent variable(s)

th1

the first threshold variable.

th2

the second threshold variable.

trim_per

trimmed percentage.

rep

nNumber of bootstrap repetitions.

plot

=1, plot; =0, do not plot.

Details

This code is the learning example for learning Hansen's econometric sample splitting. I detailed the description of each threshold stage.

Note

Original code offered by Dr. B. E.Hansen (http://www.ssc.wisc.edu/~bhansen/).

References

Hanse B. E. (2000) Sample Splitting and Threshold Estimation. Econometrica, 68, 575-603.

Examples

## Not run, becasue of bootstrap replicaiton takes time. Users may unmark # and run. 
data("dur_john")
#rep <- 500
#trim_per <- 0.15
#dep <- "gdpGrowth"
#indep <- colnames(dur_john)[c(2,3,4,5)]
#th1 <- "GDP60"
#th2 <- "Literacy"
#OUT=SMPLSplit_est(data=dur_john,dep,indep,th=th1,plot=0,h=1,nonpar=2)
#OUT$TEST
#OUT$Hypothesis
#OUT$Threshold
#stat=matrix(as.numeric(OUT$TEST),byrow = TRUE,8,2)
#colnames(stat)=c("F-Stat","P-value")
#rownames(stat)=OUT$Hypothesis
#stat

Testing for sample splitting

Description

A function for testing sample split given subsampled data.

Usage

SMPLSplit_het(data,dep,indep,th,trim_per,rep,plot)

Arguments

data

the data in either data.frame or matrix

dep

the name of dependent variable.

indep

the name(s) of independent variable(s).

th

the name of threshold variable.

trim_per

trimmed percentage.

rep

number of bootstrap repetition.

plot

=1, plot; =0, do not plot.

Details

This code tests for the presence of threshold. It generalizes the simple code of Dr. Hansen, allowing Heteroskedastic Errors (White Corrected).

Value

fstat

LM-test for no threshold.

pvalue

bootstrap P-Value.

Note

Original code offered by Dr. B. E.Hansen (http://www.ssc.wisc.edu/~bhansen/).

References

Hanse B. E. (2000) Sample Splitting and Threshold Estimation. Econometrica, 68, 575-603.

Generate a data matrix of seasonal components

Description

Generate a data matrix of seasonal components, having two pattern cycles.

Usage

SeasComponent(wts, type)

Arguments

wts

A univeriate time series with monthly or quarterly frequency.

type

Types of patterns of seasonal cycle.
="dummyCycle", generating dummy variables for the pattern of seasonal cycle, Barsky & Miron (1989)
="trgCycle", generating trigonometric variables for the pattern of seasonal cycle, Granger & Newbold (1986).

Details

This function generates data matrix for controlling the pattern of seasonal cycles. type="dummyCycle" generates DUMMY variables with season frequency. type="trgCycle" generates trigonometric pattern.

Value

A dataframe returns. Number of columns is determined by frequency.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle

References

Barsky, Robert B. and Jeffrey A. Miron (1989) The Seasonal Cycle and the Business Cycle. Journal of Political Economy, 97 (3): 503-32.
Granger, Clive William John and Newbold, Paul (1986) Forecasting Economic Time Series. 2nd edition. Published by New Milford, Connecticut, U.S.A.: Emerald Group Pub Ltd.
Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

Examples

data(inf_Q)
y=inf_Q[,2]
SeasComponent(y,type="dummyCycle")
SeasComponent(y,type="trgCycle")

Panel data of bank,2001Q1~2010Q1

Description

A quarterly panel data frame with 1000 observations on the following 7 variables, unbalanced panel data ranges from 2001Q1~2010Q1.

Usage

data("bank_income")

Format

ID

a numeric vector

Qtr

a numeric vector

preTax_Income

a numeric vector

shortRatio

a numeric vector

longRatio

a numeric vector

Current_ratio

a numeric vector

LoanDeposit_ratio

a numeric vector

Examples

data(bank_income)

Cigaretts consumption of US states

Description

Cigaretts consumption of US states

Usage

data(cigaretts)

Format

A data frame of 48 US states' cigaretts consumption

State

State abbreviation, N

Year

Year, t

Y_SALES

Cigarette sales in packs per capita, deflated by population

X1_PRICE

P=Real price per pack of cigarettes, deflated by 1983 CPI.

X2_PIMIN

Real minimum price in adjoining states per pack of cigarettes, deflated by CPI

X3_NDI

Per capita disposable income

References

Baltagi Badi H. (2005) Econometric Analysis of Panel Data. John Wiley.

Examples

data(cigaretts)
head(cigaretts)

Function for extracting components from a lm object

Description

Extract the standard error and t-stat of the a-th parameter estimate of a lm object

Usage

contts(lm, a)

Arguments

lm

lm object

a

The a-th parameter estimate of a linear model regression

Value

se.coef

The standard error of the selected coefficient

t.stat

The t-stat of the selected coefficient

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle.

References

Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

Examples

x=rnorm(100)
y=1+0.2*x+rnorm(100)
LMout=lm(y~x)
contts(LMout,1)

#$se.coef
#[1] 0.1081023

#$t.stat
#(Intercept) 
#   10.60401

Annual crime dataset of US counties

Description

Annual crime dataset of US counties

Usage

data(crime)

Format

A data frame of US counties

county

counties index, N

year

Year, t

crmrte

crime rate(crime/population)

prbarr

probability of arrest (arrests/offenses)

prbconv

probability of conviction, given arrest

prbpris

probability of a prison, given conviction

avgsen

sanction severity(average prison sentence in days )

polpc

ability of police force to detect crime(# of police per capita)

density

population density(POP/area)

taxpc

Taxpayment per capita

region

region index of county

smsa

=1 if SAMA, POP>50000; =0 else

pctmin

See Baltagi(2006) for details

wcon

See Baltagi(2006) for details

wtuc

See Baltagi(2006) for details

wtrd

See Baltagi(2006) for details

wfir

See Baltagi(2006) for details

wser

See Baltagi(2006) for details

wmfg

See Baltagi(2006) for details

wfed

See Baltagi(2006) for details

wsta

See Baltagi(2006) for details

wloc

See Baltagi(2006) for details

mix

See Baltagi(2006) for details

pctymle

See Baltagi(2006) for details

References

Baltagi Badi H. (2005) Econometric Analysis of Panel Data. John Wiley. Baltagi Badi H. (2006) Estimating an Economic Model of Crime Using Panel Data from North Carolina. J.of Applied Econometrics 21: 543¡V547.

The cross-country growth data in Durlauf and Johnson(1995)

Description

The Durlauf-Johnson data manupipulated by Hansen(2000),excluding missing variables and oil states

Usage

data(dur_john)

Format

A data frame with 19 countries

gdpGrowth

Economic growth measured by GDP of 1960 and 1985

logGDP60

log Per capita GDP in 1960

Inv_GDP

Average ratio of investment (including Government Investment) to GDP from 1960 to 1985

popGrowth

Average growth rate of working-age population 1960 to 1985

School

Average fraction of working-age population enrolled in secondary school from 1960 to 1985

GDP60

Per capita GDP in 1960

Literacy

fraction of the population over 15 years old that is able to read and write in 1960

Details

Steven N. Durlauf and Paul A. Johnson, "Multiple Regimes and Cross-Country Growth Behavior," Journal of Applied Econometrics, Vol. 10, No. 4, 1995, 365-384.

Examples

data(dur_john)
head(dur_john)

Generate the HEGY regressors.

Description

This function generates the level regresors in HEGY regression, without differenced lag terms.

Usage

hegy.reg(wts)

Arguments

wts

Univariate time series, with a possibly seasonal stochastic trend

Details

This function automatically identifies the frequency of time series data, and generate necessary level components as described in Eq.(3.7) of Hylleberg et. al (1990).

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle

References

Hylleberg, S., Engle, R.F., Granger, C.W.J., and Yoo, B.S.(1990) Seasonal integration and cointegration. Journal of Econometrics,44, 215-238.
Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

Examples

data(inf_Q)
y=inf_Q[,1]
hegy.reg(y)

Specification test for panel glm models

Description

This function performs Hausman specification test for panel glm.

Usage

htest_pglm(RE, FE, re.method)

Arguments

RE

Random effect objects. Support pglm, glmer, glmmTMB

FE

Fixed effect objects.

re.method

Method that used to estimate the random effect estimation, in addition to "pglm", it also supports "glmmTMB" of package glmmTMB, and "glmer" of package lme4.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from phtest() of plm

References

Hausman J.A. (1978). Specification Tests in Econometrics. Econometrica, 46, 1251-1271.

Examples

data(ship)
library(pglm)
Eq1="accident  ~  op_75_79+co_65_69+co_70_74+co_75_79"
FE.pois <- pglm(Eq1,data=ship,family = "poisson",model = "within",index = 'ship',R=10)

RE.pois <- pglm::pglm(Eq1,data=ship,family = "poisson", model = "random", index = 'ship')

## Hausman test
htest_pglm(RE=RE.pois,  FE=FE.pois, re.method="pglm")

Eq2=accident  ~  op_75_79+co_65_69+co_70_74+co_75_79 + (1 | ship)
re.glmmTMB=glmmTMB::glmmTMB(Eq2,data=ship, family="poisson")

## Hausman test
htest_pglm(RE=re.glmmTMB,  FE=FE.pois, re.method="glmmTMB")

Monthly inflation time series of 19 countries

Description

Monthly inflation time series of 19 countries, 1984.1~2011.3

Usage

data(inf19)

Format

A data frame with 19 countries

AUSTRIA

inflation of Austria

BELGIUM

inflation of Belgium

CANADA

inflation of Canada

DENMARK

inflation of Denmark

FINLAND

inflation of Finland

FRANCE

inflation of France

GREECE

inflation of Greece

ICELAND

inflation of Iceland

ITALY

inflation of Italy

JAPAN

inflation of Japan

LUXEMBOURG

inflation of Luxembourg

NETHERLANDS

inflation of Netherlands

NORWAY

inflation of Norway

PORTUGAL

inflation of Portugal

SPAIN

inflation of Spain

SWEDEN

inflation of Sweden

SWITZERLAND

inflation of Switzerland

UK

inflation of UK

USA

inflation of USA

Details

Monthly CIP, seasonaly differenced of log CPI of 19 countries

Examples

data(inf19)
head(inf19)

Monthly inflation time series of 20 countries

Description

Monthly inflation time series of 19 countries, 1971.1~2011.12

Usage

data(inf_M)

Format

A data frame with 20 countries

AUSTRALIA

inflation of Austrlia

AUSTRIA

inflation of Austria

BELGIUM

inflation of Belgium

CANADA

inflation of Canada

DENMARK

inflation of Denmark

FINLAND

inflation of Finland

FRANCE

inflation of France

GREECE

inflation of Greece

ICELAND

inflation of Iceland

ITALY

inflation of Italy

JAPAN

inflation of Japan

LUXEMBOURG

inflation of Luxembourg

NETHERLANDS

inflation of Netherlands

NORWAY

inflation of Norway

PORTUGAL

inflation of Portugal

SPAIN

inflation of Spain

SWEDEN

inflation of Sweden

SWITZERLAND

inflation of Switzerland

UK

inflation of UK

USA

inflation of USA

Details

Monthly CIP, seasonaly differenced of log CPI of 20 countries

Examples

data(inf_M)
head(inf_M)

Quarterly inflation time series of 20 countries

Description

Quarterly inflation time series of 19 countries, 1971Q1~2014Q4

Usage

data(inf_Q)

Format

A data frame with 19 countries

AUSTRLIA

inflation of Austrlia

AUSTRIA

inflation of Austria

BELGIUM

inflation of Belgium

CANADA

inflation of Canada

DENMARK

inflation of Denmark

FINLAND

inflation of Finland

FRANCE

inflation of France

GREECE

inflation of Greece

ICELAND

inflation of Iceland

ITALY

inflation of Italy

JAPAN

inflation of Japan

LUXEMBOURG

inflation of Luxembourg

NETHERLANDS

inflation of Netherlands

NORWAY

inflation of Norway

PORTUGAL

inflation of Portugal

SPAIN

inflation of Spain

SWEDEN

inflation of Sweden

SWITZERLAND

inflation of Switzerland

UK

inflation of UK

USA

inflation of USA

Details

Quarterly CIP, seasonaly differenced of log CPI of 20 countries

Examples

data(inf_Q)
head(inf_Q)

Extracting critical value and p-value from Table 1 of Hylleberg et. al (1990)

Description

Hylleberg et. al (1990,pp.226-227) offer simulated critical values for seasonal unitr to test. interpolpval() is an internal call and should not be used independently.

Usage

interpolpval(code, stat, N, swarn = TRUE)

Arguments

code

Type of HEGY model, this will be automatically identified.

stat

Empirical test statistics.

N

Sample size calculating stat above.

swarn

Logical. Whether the warning message for negative p-value will be returned? The default is TRUE.

Value

table

Table for critical value and p-value.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle

References

Hylleberg, S., Engle, R.F., Granger, C.W.J., and Yoo, B.S.(1990) Seasonal integration and cointegration. Journal of Econometrics,44, 215-238.
Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

investment data of 565 listed companies, 1973-1987

Description

investment data of 565 listed companies, 1973-1987, from Hansen's example

Usage

data(invest)

Format

A pooled data frame

invest[,1]

investment/assets

invest[,2]

Tobin's Q

invest[,3]

cash-flow/assets

invest[,4]

debt/assets

Details

This is a pooled data frame, without date (T) and cross-section(N) ID columes

Examples

#data(invest)
#head(invest)

IPS-HEGY seasonal unit root test in panel data, Otero et al.(2007).

Description

This function performs panel data-based HEGY seasonal unit root test, the asymptotics is based upon Otero et al.(2007).

Usage

ipsHEGY(data, itsd, Sel, pmax, CIPS = TRUE)

Arguments

data

Panel data, T by N

itsd

Options for c(i,t,sd).
i=1, intercept;=0 no intercept.
t=1, trend;=0 no deterministic trend.
sd=1, season dummy 1:(s-1);=0 no.

Sel

Selection of lags, having three options: "signf","bic","aic".

pmax

Maximum number of lags for searching optimal criteria.

CIPS

Logical. If TRUE, using Pesaran(2007) to account for cross-section correlation. The default is TRUE.

Details

Mode for selectlags has three options, AIC and BIC use R built-in functions for linear model and their meanings are popular and straightforward. "signf" includes only statistically significant lags, and statistically insignificant lags are dropped from regressors. That is, once you select this option, lags of your model may not be continuous.
The critical values for panel HEGY are standard normal for individual t-ratios, however,you need to perform simulation for the critical values of F joint test, at pdR 1.3. To this end, you are encouraged to work this out for yourself: using arima.sim() to sample seasonal time series with unit root (1-order difference) and obtain their statistics under the null using ipsHEGY(), then it is streightforward to obtain critical values.
Otero et al. (2007) provide critical values for quarterly frequency.

Value

P_HEGY

Panel HEGY statistics.

U_HEGY

Idividual HEGY statistics of N units.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, College of Management, National Taiwan Normal University.

References

Otero, J., Smith, J., and Giulietti, M. (2005) Testing for seasonal unit roots in heterogeneous panels. Economics Letters, 86, 229-235.
Otero, J., Smith, J., and Giulietti, M. (2007) Testing for seasonal unit roots in heterogeneous panels in the presence of cross section dependence. Economics Letters, 86, 179-184.
Pesaran M. Hashem (2007) A simple panel unit root test in the presence of cross-section dependence. Journal of Applied Econometrics, 22, 265-312.

Examples

data(inf_Q)
dataz<-inf_Q
itsd<-c(1,0,c(1:3))
#Seasonal dummy only takes quarters 1:3, 
#becasue of the presence of common intercept.
Sel<-"bic" # "aic","bic", "signf".
pmax<-12

OUT<-ipsHEGY(dataz,itsd,Sel,pmax,CIPS=FALSE)
OUT$P_HEGY
OUT$U_HEGY

# Simulation of critical values

Select the optimal number of lags, given criteria

Description

Determine the optimal number of lags for dynamic regression

Usage

lagSelect(y, maxp, ic)

Arguments

y

A univariate time series data

maxp

the max number of lags

ic

Information criteria, either "AIC" or "BIC"

Details

Information criteria "AIC" and "BIC" use the R built-in functions.

Value

It returns an integer, indicating the optimal lags

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, College of Management, National Taiwan Normal University.

Examples

#library(pdR)
#data(inf19)
#y<-inf19[,1]
#lagSelect(y,maxp=25,ic="BIC")

Function for looking up tabulated critical values and associated p-values of HEGY test.

Description

Function for looking up tabulated critical values and associated p-values,Hylleberg et. al (1990, Table 1a and Table 1b).

Usage

lookupCVtable(code)

Arguments

code

Type of HEGY model, this will be automatically identified.

Value

table

Table for critical value and p-value.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle

References

Hylleberg, S., Engle, R.F., Granger, C.W.J., and Yoo, B.S.(1990) Seasonal integration and cointegration. Journal of Econometrics,44, 215-238.
Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

Estimate specified panel threshold model

Description

This function is the main function estimating threshold regression for function ptm()

Usage

model(r, trim, rep, it, qq1, cf, xt, ct, thresh, tt, qn1, n, qn, cc, yt, ty, k)

Arguments

r

vector of threshold estimate(s).

trim

value of trimmed percentage.

rep

number bootstrap repetition.

it

number of regime during computation, used in a for loop.

qq1

defined parameter.

cf

special declaration, e.g. lag().

xt

regime independent variables.

ct

trace of regime dependent variables.

thresh

threshold variable.

tt

length of time period.

qn1

as defined by nrow(qq1).

n

number of cross-section units.

qn

number of quantiles to examine.

cc

as defined by 2*log(1-sqrt(conf_lev)).

yt

vectorized dependent variable.

ty

trace of yt.

k

number of regime-independent independent variables.

Note

Original code offered by Dr. B. E.Hansen (http://www.ssc.wisc.edu/~bhansen/).

References

Hanse B. E. (1999) Threshold effects in non-dynamic panels: Estimation, testing and inference. Journal of Econometrics,93, 345-368.

Panel unit root test of Chang(2002)

Description

Compute the panel unit root test statistic of Chang(2002).

Usage

pIGF(datamat, maxp, ic, spec)

Arguments

datamat

T by N panel data.T is the time length,N is the number of cross-section units.

maxp

the max number of lags

ic

Information criteria, either "AIC" or "BIC".

spec

model specification.
=0, no intercept and trend.
=1, intercept only.
=2, intercept and trend.

Details

This function estimates the panel unit root test based on univariate instrument generating function of (Chang,2002).

Value

panel.tstat

panel IGF test statistics

pvalue

P-value of the panel.tstat

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, College of Management, National Taiwan Normal University.

References

Chang,Y.(2002) Nonlinear IV Unit Root Tests in Panels with Cross-Sectional Dependency. Journal of Econometrics, 110, 261-292.

Examples

data(inf19)
datam <- inf19
pIGF(datam,maxp=25,ic="BIC",spec=2)

Productivity data of 48 US state,1970-1986

Description

Gross state production data

Usage

data(productivity)

Format

A data frame with US production

state

US state index, 1-48

year

Year index

y_gsp

Gross state product

x1_hwy

Expediture of public utility- highway construction

x2_water

Expediture of public utility- water

x3_other

Expediture of others

x4_private

Private consumption of each state

x5_emp

Employment rate of each state

x6_unemp

Unemployment rate of each state

Examples

data(productivity)
head(productivity)

Threshold specification of panel data

Description

A generalized specification for estimating panel threshold model.

Usage

ptm(dep, ind1, ind2, d, bootn, trimn, qn, conf_lev, t, n)

Arguments

dep

Dependent variable

ind1

Independent variables: regime dependent

ind2

Independent variables:regime independent

d

Threshold variable

bootn

Vector of bootstrap repetition

trimn

Vector of trimmed percentage

qn

Number of quantiles to examine

conf_lev

Confidence level

t

Length of time period

n

Number of cross-section units

Details

This code fits only balanced panel data. It generalizes the simple code of Dr. Hansen (http://www.ssc.wisc.edu/~bhansen/), allowing multiple (more-than-one) regime-dependent (ind1) variables. We generalize the original code to better fit general need of threshold modeling in panel data.
bootn and trimn are vector of 3 by 1, indicating numbers of three corresponding regimes.
This version corrects a slight error incurred by argument max_lag, which is used by Hansen to arrange investment data via lags. In this package, users manipulate data to fit personal research to ptm(), hence this argument is omitted lest should degree of freedom will suffer a loss of N.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, College of Management, National Taiwan Normal University.

References

Hansen B. E. (1999) Threshold effects in non-dynamic panels: Estimation, testing and inference. Journal of Econometrics,93, 345-368.

Examples

# library(pdR)
#data(invest)
#dat<-invest[1:1500,]    # subsetting the first 1500 obs., #for simplicity
#t <- 15            #Length of time period
#nt <- nrow(dat)
#n <- nt/t           # number of cross-section units

#dep<- as.matrix(dat[,1])       # investment/assets
#th1<- as.matrix(dat[,2])  #Tobin's Q
#th2<- as.matrix(dat[,3])  # cash-flow/assets
#ind1<- cbind(th1,th2)  #regime-dep covariates 
#d <- as.matrix(dat[,4])      # Threshold variable      
#ind2 <- cbind((th1^2),(th1^3),(th1*d)) # regime-indep covariates:
#bootn<-c(100,200,300)  # bootstrapping replications for each  threshold esitmation
#trimn<-c(0.05,0.05,0.05)   #trimmed percentage for each threshold  esitmation

#qn<-400
#conf_lev<-0.95

#Output=ptm(dep,ind1,ind2,d,bootn,trimn,qn,conf_lev,t,n)
#Output[[1]] #Formatted output of 1st threshold, 2 regimes
#Output[[2]] #Formatted output of 2nd threshold, 3 regimes
#Output[[3]] #Formatted output of 3rd threshold, 4 regimes

# In the output, the Regime-dependent Coefficients matrix
# is, from top to bottom, regime-wise.

A subroutine for model()

Description

This function is a subroutine for model(), estimation procedure.

Usage

r_est(y, r, trim, tt, qq1, qn1, qn, n, cf, xt, ct, thresh)

Arguments

y

vector of dependent variable.

r

numer of regime.

trim

value of trimmed percentage.

tt

length of time period.

qq1

parameter defined by as.matrix(unique(thresh)[floor(sq*nrow(as.matrix(sort(unique(thresh)))))]).

qn1

as defined by nrow(qq1).

qn

number of quantiles to examine.

n

parameter of cross-section units.

cf

special declaration, e.g. lag().

xt

regime independent variables.

ct

trace of regime dependent variables.

thresh

threshold variable.

References

Hanse B. E. (1999) Threshold effects in non-dynamic panels: Estimation, testing and inference. Journal of Econometrics,93, 345-368.

Original code from Dr. Hansen (http://www.ssc.wisc.edu/~bhansen/).

Returns a data.frame of sequential lag matrix.

Description

ret() is similar to embed(), but returns a data.frame specified with colnames, not matrix.

Usage

ret(wts, k)

Arguments

wts

Univariate time series.

k

k-1 lagged terms.

Details

ret() is similar to embed(), but returns a data.frame with colnames, not matrix. Moreover, unlike embed(),ret() fills lagged cells with NA, instead of trimming them.

Value

A T by k dataframe returns. If you need 2 lags, you have to specify k=3, then it returns a dataframe with T by 3 dataframe, the first column is lag0.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle

References

Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

Examples

data(inf_Q)
y=inf_Q[,2]
ret(y,3)

Selection of lags.

Description

Lagged coefficient estimates are kept if they meet the inequality condition of AIC or BIC.

Usage

selPabic(lmdet, type, Pmax = NULL)

Arguments

lmdet

Object of lm()

type

Take the value of "aic" or "bic".

Pmax

The maximum number of lag orders.

Details

This is an internal function used for HEGY.test(). Beginning with pamx, the lag order will be drop if its inclusion worsens the minimum condition. Hence, they may not be a regular sequence. For example, for pmax=10, the selected lags may look like (1,4,5,8,9), rather than 1,2,3,...10.

Value

This function returns the lag orders.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle

References

Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

Examples

data(inf_Q)
y=inf_Q[,1]
hegy.out<-HEGY.test(wts=y, itsd=c(1,0,c(1:3)),regvar=0, selectlags=list(mode="aic", Pmax=12))
hegy.out$lagsorder
hegy.out$lagcoefs

Selection of lags.

Description

Lagged coefficient estimates are kept if they are statistically significant

Usage

selPsignf(lmdet, cvref = 1.65, Pmax = NULL)

Arguments

lmdet

Object of lm()

cvref

Reference of critical values, the default is 1.65.

Pmax

The maximum number of lag orders.

Details

This is an internal function used for HEGY.test(). Beginning with pamx, the lag order will be kept if it is statistically significant. Hence, the lag orders may not be a regular sequence. For example, for pmax=10, the seelcted lags may look like (1,4,5,8,9), rather than 1,2,3,...10.

Value

This function returns the lag orders.

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>, modifed from Javier Lopez-de-Lacalle

References

Javier Lopez-de-Lacalle in https://github.com/cran/uroot/blob/master/R/hegy.R

Examples

data(inf_Q)
y=inf_Q[,1]
hegy.out<-HEGY.test(wts=y, itsd=c(1,0,c(1:3)),regvar=0, selectlags=list(mode="signf", Pmax=12))
hegy.out$lagsorder
hegy.out$lagcoefs

Panel data on the number of ship accidents

Description

Panel data on the number of ship accidents, McCullagh and Nelder(1989)

Usage

data("ship")

Format

accident

the number of ship accidents

ship

Ship iD

service

the number of months in service

op_75_79

the operating period between 1975 and 1979

co_65_69

consecutive construction periods of 5 years

co_70_74

consecutive construction periods of 5 years

co_75_79

consecutive construction periods of 5 years

yr_con

years of construction

yr_op

years of operation

References

McCullagh, P., and J. A. Nelder (1989) Generalized Linear Models. 2nd ed. London: Chapman and Hall/CRC.

Examples

data(ship)

a subroutine of model()

Description

SSE calculation

Usage

sse_calc(y, x)

Arguments

This function is a sub-routine for model(), calculating SSE of each regression

y

vector of dependent variable.

x

matrix of independent variables.

References

Hanse B. E. (1999) Threshold effects in non-dynamic panels: Estimation, testing and inference. Journal of Econometrics,93, 345-368.

Original code from Dr. Hansen (http://www.ssc.wisc.edu/~bhansen/).

Compute the resursive mean

Description

Compute the resursive mean of each series

Usage

 tbar(x)

Arguments

x

A univariate time series data

Details

This function computes the resursive mean

Author(s)

Ho Tsung-wu <tsungwu@ntnu.edu.tw>

Examples

data(inf19)
y <- inf19[,1]
tbar(y)

a subroutine calculating SSE

Description

This function is a sub-routine for model(), calculating SSE of each threshold regression.

Usage

thr_sse(y, q, r, cf, xt, ct, thresh, tt, n)

Arguments

y

parameter.

q

qq1 in model().

r

parameter.

cf

as defined in model().

xt

as defined in model().

ct

as defined in model().

thresh

as defined in model().

tt

as defined in model().

n

as defined in model().

References

Hanse B. E. (1999) Threshold effects in non-dynamic panels: Estimation, testing and inference. Journal of Econometrics,93, 345-368.

Original code from Dr. Hansen (http://www.ssc.wisc.edu/~bhansen/).

A sub-routine calculating trace

Description

Estimation of trace.

Usage

 tr(y, tt, n)

Arguments

This function is a sub-routine for model(), calculating trace of matrix

y

data vector.

tt

time period length.

n

number of cross-section units.

References

Hanse B. E. (1999) Threshold effects in non-dynamic panels: Estimation, testing and inference. Journal of Econometrics,93, 345-368.

Original code from Dr. Hansen (http://www.ssc.wisc.edu/~bhansen/).