Version: 1.0
Date: 2017-12-13
Title: Pooled Mean Group Estimation of Dynamic Heterogenous Panels
Author: Piotr Zientara [aut], Lech Kujawski [aut, cre]
Maintainer: Lech Kujawski <lech.kujawski@ug.edu.pl>
Depends: R (≥ 3.2.3)
Description: Calculates the pooled mean group (PMG) estimator for dynamic panel data models, as described by Pesaran, Shin and Smith (1999) <doi:10.1080/01621459.1999.10474156>.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]
URL: https://www.r-project.org
NeedsCompilation: no
Packaged: 2017-12-14 08:45:00 UTC; Marcin
Repository: CRAN
Date/Publication: 2017-12-14 13:08:28 UTC

Pooled Mean Group Estimation of Dynamic Heterogenous Panels

Description

Calculates the pool mean group (PMG) estimator for dynamic panel data models, as described by Pesaran, Shin and Smith (1999) <doi:10.1080/01621459.1999.10474156>. This estimator enables the intercepts, short-run coefficient and error variances to differ freely across groups, but restricts the long-run coefficients to being equal. Additionally, it allows the numbers of time series observations to differ freely across groups. This software also performs diagnostic tests of error terms, such as autocorrelation, heteroscedasticity and normality. Calculates the pooled mean group (PMG) estimator for dynamic panel data models, as described by Pesaran, Shin and Smith (1999) <doi:10.1080/01621459.1999.10474156>.

Details

The DESCRIPTION file:

Package: PooledMeanGroup
Version: 1.0
Date: 2017-12-13
Title: Pooled Mean Group Estimation of Dynamic Heterogenous Panels
Authors@R: c(person("Piotr", "Zientara", role="aut", email="zientara@fest.pl"), person("Lech", "Kujawski", role=c("aut", "cre"), email = "lech.kujawski@ug.edu.pl"))
Author: Piotr Zientara [aut], Lech Kujawski [aut, cre]
Maintainer: Lech Kujawski <lech.kujawski@ug.edu.pl>
Depends: R (>= 3.2.3)
Description: Calculates the pooled mean group (PMG) estimator for dynamic panel data models, as described by Pesaran, Shin and Smith (1999) <doi:10.1080/01621459.1999.10474156>.
License: GPL (>= 2)
URL: https://www.r-project.org

Index of help topics: 

BGtest                  BGtest
ConoverMulti            ConoverMulti
DataExp                 DataExp
DiffPanel               DiffPanel
GQtest                  GQtest
JBtest                  JBtest
LagPanel                LagPanel
PMG                     PMG
PanelNaOmit             PanelNaOmit
PooledMeanGroup-package
                        Pooled Mean Group Estimation of Dynamic
                        Heterogenous Panels
optimPMG                optimPMG

Author(s)

Lech Kujawski, Piotr Zientara Piotr Zientara [aut], Lech Kujawski [aut, cre]

Maintainer: Lech Kujawski <lech.kujawski@ug.edu.pl>

References

Pesaran, Shin and Smith (1999) <doi:10.1080/01621459.1999.10474156>

BGtest

Description

Tests autocorrelation between the current and lagged residuals. The test is a joint test of the first P autocorrelations

Usage

BGtest(residuals, explvariab, acor.ord)

Arguments

residuals

residuals for group i

explvariab

explanatory variables (regressors) for group i

acor.ord

order of tested autocorrelations

Details

Calculates statistics and probs of the Breusch-Godfrey autocorrelation test (with two variants: chi-squared and F

Value

Chi-squared and F statistics with probs

Author(s)

Lech Kujawski, Piotr Zientara

Examples

# creating artificial variables
x1=rnorm(30,0,1)
x2=rnorm(30,0,1)
e=rnorm(30,0,0.2)
y=1+2*x1+3*x2+e
# any model
model=lm(y~x1+x2)
# BGtest
ExpBGtest=BGtest(residuals=resid(model), explvariab=cbind(x1,x2), acor.ord=4)
ExpBGtest

ConoverMulti

Description

Tests for homoscedasticity among subsamples k within a particular group i (note that the Conover test is a non-parametric test)

Usage

ConoverMulti(residuals, subsample)

Arguments

residuals

residuals for group i

subsample

the vector of c(s1i, s2i,..., ski), where s1i+s2i+...+ski=Ti (i.e., the vector divides a particular group i into subsamples)

Details

Calculates chi-squared statistic with a prob

Value

Chi-squared statistic with a prob

Author(s)

Lech Kujawski, Piotr Zientara

Examples

# creating artificial variables
x1=rnorm(30,0,1)
x2=rnorm(30,0,1)
e=rnorm(30,0,0.2)
y=1+2*x1+3*x2+e
# any model
model=lm(y~x1+x2)
# ConoverMulti
ExpConoverMulti=ConoverMulti(residuals=resid(model), subsample=c(10,10,10))
ExpConoverMulti

DataExp

Description

A dataset in the form of stacked time-series. Quarterly data cover nine countries (Poland, Bulgaria, the Czech Republic, Hungary, Latvia, Lithuania, Romania, Slovakia, Slovenia; the numbers denoting particular countries form a series i=1,2,...,9) from 2005q2 to 2013q4. The dataset contains the following (below) variables

Usage

data("DataExp")

Format

A data frame with 315 observations on the following 16 variables.

y10

a numeric vector

y10spread

a numeric vector

riskavers

a numeric vector

debt

a numeric vector

deficit

a numeric vector

openess

a numeric vector

cpi

a numeric vector

growth

a numeric vector

crisk

a numeric vector

urate

a numeric vector

iip

a numeric vector

iipnetto

a numeric vector

cagdp

a numeric vector

caresvs

a numeric vector

cds

a numeric vector

bidask

a numeric vector

Examples

data(DataExp)

DiffPanel

Description

Calculates first differences of a particular variable from a panel data set

Usage

DiffPanel(variable, quantity)

Arguments

variable

a particular variable from a panel data set in the form of stacked time series; in practice; a selected singular column from a panel data set

quantity

a vector of the number of time series observations in each group; in practice, it takes the form c(T1,...Tn) since the PMG allows the numbers of time series observations to differ freely across groups (if the number of time series observations in each group is the same, then c(T,...,T) and T=T1=T2=...=Tn

Details

Calculates first differences of a particular variable from a panel data set in order to bring it to stationarity. Preserves the original dimension of time series observations in each group, completing data lost due to differentiating by inserting "NA"

Value

First differences of a particular variable from a panel data set

Author(s)

Lech Kujawski, Piotr Zientara

Examples

# first import DataExp, i=1...9, T1=T2=...T9=35
data(DataExp)
DataExp[1:5,]
# then execute DiffPanel
y10=data.frame(y10=DataExp[,1], row.names=row.names(DataExp))
dy10=DiffPanel(variable=y10, quantity=rep(35,9))
diip=DiffPanel(variable=DataExp[,11], quantity=rep(35,9))
cbind(y10,dy10,diip)[1:5,]

GQtest

Description

Tests for homoscedasticity

Usage

GQtest(residuals, subsample, nep)

Arguments

residuals

residuals for group i

subsample

the vector of c(s1i, s2i), where s1i=Ti/2 and s2i=s1i+1 (if Ti is EVEN) or where s1i=Ti/2-0.5 and s2i=s1i+1 (if Ti is UNEVEN)

nep

the number of estimated parameters for group i

Details

Calculates F statistic with a prob

Value

F statistic with a prob

Author(s)

Lech Kujawski, Piotr Zientara

Examples

# creating artificial variables
x1=rnorm(30,0,1)
x2=rnorm(30,0,1)
e=rnorm(30,0,0.2)
y=1+2*x1+3*x2+e
# any model
model=lm(y~x1+x2)
#BGtest
ExpGQtest=GQtest(residuals=resid(model), subsample=c(15,16), nep=3)
ExpGQtest

JBtest

Description

Tests for normality

Usage

JBtest(residuals)

Arguments

residuals

residuals for group i

Details

Calculates chi-squared statistic with a prob

Value

Chi-squared statistic with a prob

Author(s)

Lech Kujawski, Piotr Zientara

Examples

# creating artificial variables
x1=rnorm(30,0,1)
x2=rnorm(30,0,1)
e=rnorm(30,0,0.2)
y=1+2*x1+3*x2+e
# any model
model=lm(y~x1+x2)
#JBtest
ExpJBtest=JBtest(residuals=resid(model))
ExpJBtest

LagPanel

Description

Provides the first lag of a particular variable from a panel data set

Usage

LagPanel(variable, quantity)

Arguments

variable

a particular variable from a panel data set in the form of stacked time series; in practice; a selected singular column from a panel data set

quantity

a vector of the number of time series observations in each group; in practice, it takes the form c(T1,...Tn) since the PMG allows the numbers of time series observations to differ freely across groups (if the number of time series observations in each group is the same, then c(T,...,T) and T=T1=T2=...=Tn

Details

Provides the first lag of a particular variable from a panel data set. Preserves the original dimension of time series observations in each group, completing data lost due to lagging by inserting "NA"

Value

A lagged particular variable from a panel data set

Author(s)

Lech Kujawski, Piotr Zientara

Examples

# first import DataExp, i=1...9, T1=T2=...T9=35
data(DataExp)
DataExp[1:5,]
# then execute LagPanel
y10=data.frame(y10=DataExp[,1], row.names=row.names(DataExp))
ly10=LagPanel(variable=y10, quantity=rep(35,9))
ldebt=LagPanel(variable=DataExp[,4], quantity=rep(35,9))
cbind(y10,ly10,ldebt)[1:5,]

PMG

Description

Having particular long-run parameters (exp. start values) estimates parameters of short-run relationships as well as standard errors of estimations, Student's t-distribution type statistics, probs, confidence intervals. Also performs diagnostic tests of error terms, such as autocorrelation, heteroscedasticity and normality

Usage

PMG(paramTeta, vecSR, vecLR, dataset, quantity, const)

Arguments

paramTeta

the vector of parameters of long-run relationships, as outlined in Equation 7 (Pesaran, Shin and Smith, 1999)

vecSR

a list of vectors containing the column numbers of variables in short-run relationships for each group (alternatively a list of vectors containing the variables names instead of column numbers). In each vector of the list the first number must indicate dy (i.e., the dependant variable)

vecLR

a vector containing the column numbers of variables in long-run relationships (alternatively a vector containing the variables names instead of column numbers). The first number must indicate ly (i.e., the lagged dependant variable)

dataset

a panel data set in the form of stacked time series, containing variables of long-run and short-run relationships (i.e., including differentiated and lagged variables)

quantity

a vector of the number of time series observations in each group; in practice, it takes the form c(T1,...,Tn) since the PMG allows the numbers of time series observations to differ freely across groups (if the number of time series observations in each group is the same, then c(T,...,T) and T=T1=T2=...=Tn

const

logical. If TRUE (the default value), the intercept term is added to the model (i.e., to the short-run relationship)

Details

Having particular long-run parameters estimates parameters of short-run relationships. Also estimates the information matrix as well as standard errors of estimations, as indicated in Equation 13 (Pesaran, Shin and Smith, 1999). Calculates Student's t-distribution type statistics, probs and confidence intervals. Also performs diagnostic tests of error terms, such as the Breusch-Godfrey autocorrelation test, the Goldfeld-Quandt heteroscedasticity test and the Conover nonparametric test of homogeneity of variance and the Jarque-Bera normality test

Value

$LogL

the concentrated log-likelihood function

$LR

parameters of long-run relationships

$SR

the estimated parameters of short-run relationships

$DiagTests

results of diagnostic tests

$residuals

residuals

Author(s)

Lech Kujawski, Piotr Zientara

References

Pesaran, Shin and Smith (1999) <doi:10.1080/01621459.1999.10474156>

Examples

# first import DataExp, i=1...9, T1=T2=...T9=35
data(DataExp)
DataExp[1:5,]
# then prepare lags and diffs using LagPanel and DiffPanel
y10=data.frame(y10=DataExp[,1], row.names=row.names(DataExp))
cpi=data.frame(cpi=DataExp[,7], row.names=row.names(DataExp))
dy10=DiffPanel(variable=y10, quantity=rep(35,9))
dopeness=DiffPanel(variable=DataExp[,6], quantity=rep(35,9))
ly10=LagPanel(variable=y10, quantity=rep(35,9))
diip=DiffPanel(variable=DataExp[,11], quantity=rep(35,9))
dcrisk=DiffPanel(variable=DataExp[,9], quantity=rep(35,9))
ldcrisk=LagPanel(variable=dcrisk, quantity=rep(35,9))
dcpi=DiffPanel(variable=DataExp[,7], quantity=rep(35,9))
ddcpi=DiffPanel(variable=dcpi, quantity=rep(35,9))
ldebt=LagPanel(variable=DataExp[,4], quantity=rep(35,9))
# create homogenous preliminary dataset (containing "NA" as a result of DiffPanel, LagPanel)
dataPanel=cbind(y10, dy10, ly10, DataExp[,6], dopeness, diip,
DataExp[,11], ldcrisk, DataExp[,9], ddcpi, DataExp[,7])
dataPanel=data.frame(dataPanel)
names(dataPanel)=c("y10", "dy10", "ly10", "openess", "dopeness", "diip",
"iip", "ldcrisk", "crisk", "ddcpi", "cpi")
dataPanel[1:5,]
# prepare dataset and quantity for PMG or optimPMG functions using PanelNaOmit
dataPanel=PanelNaOmit(dataset=dataPanel, quantity=rep(35,9))
dataPanel$dataset[1:5,]
dataPanel$quantity
# PMG
PmgExp=PMG(
paramTeta=c(-14.22768, -23.84427, -0.75717, 27.57753),
vecSR=list(SR1=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR2=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR3=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR4=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR5=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR6=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR7=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR8=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR9=c("dy10","dopeness","diip","ldcrisk","ddcpi")),
vecLR=c("ly10","openess","iip","crisk","cpi"),
dataset=dataPanel$dataset,
quantity=dataPanel$quantity,
const=TRUE)
PmgExp

PanelNaOmit

Description

Prepares a panel data set for further calculations by eliminating "NA" and modifying quantity or a vector of the number of time series observations in each group

Usage

PanelNaOmit(dataset, quantity)

Arguments

dataset

a panel data set in the form of stacked time series, containing variables of long-run and short-run relationships (i.e., including differentiated and lagged variables from DiffPanel or LagPanel)

quantity

a vector of the number of time series observations in each group; in practice, it takes the form c(T1,...Tn) since the PMG allows the numbers of time series observations to differ freely across groups (if the number of time series observations in each group is the same, then c(T,...,T) and T=T1=T2=...=Tn

Details

Eliminates "NA" and modifies quantity or a vector of the number of time series observations in each group

Value

$dataset

panel data set for further calculations modified by eliminating "NA"

$quantity

modified vector of the number of time series observations in each group

Author(s)

Lech Kujawski, Piotr Zientara

Examples

# first import DataExp, i=1...9, T1=T2=...T9=35
data(DataExp)
DataExp[1:5,]
# then prepare lags and diffs using LagPanel and DiffPanel
y10=data.frame(y10=DataExp[,1], row.names=row.names(DataExp))
cpi=data.frame(cpi=DataExp[,7], row.names=row.names(DataExp))
dy10=DiffPanel(variable=y10, quantity=rep(35,9))
dopeness=DiffPanel(variable=DataExp[,6], quantity=rep(35,9))
ly10=LagPanel(variable=y10, quantity=rep(35,9))
diip=DiffPanel(variable=DataExp[,11], quantity=rep(35,9))
dcrisk=DiffPanel(variable=DataExp[,9], quantity=rep(35,9))
ldcrisk=LagPanel(variable=dcrisk, quantity=rep(35,9))
dcpi=DiffPanel(variable=DataExp[,7], quantity=rep(35,9))
ddcpi=DiffPanel(variable=dcpi, quantity=rep(35,9))
ldebt=LagPanel(variable=DataExp[,4], quantity=rep(35,9))
# create homogenous preliminary dataset (containing "NA") after DiffPanel, LagPanel
dataPanel=cbind(y10, dy10, ly10, DataExp[,6], dopeness, diip,
DataExp[,11], ldcrisk, DataExp[,9], ddcpi, DataExp[,7])
dataPanel=data.frame(dataPanel)
names(dataPanel)=c("y10", "dy10", "ly10", "openess", "dopeness", "diip",
"iip", "ldcrisk", "crisk", "ddcpi", "cpi")
dataPanel[1:5,]
# prepare dataset and quantity for PMG or optimPMG functions using PanelNaOmit
dataPanel=PanelNaOmit(dataset=dataPanel, quantity=rep(35,9))
dataPanel$dataset[1:5,]
dataPanel$quantity

optimPMG

Description

Estimates parameters of long-run and short-run relationships. Makes use of a "back-substitution" algorithm, as described by Pesaran, Shin and Smith (1999). Also estimates the information matrix as well as standard errors of estimations, as indicated in Equation 13 (Pesaran, Shin and Smith, 1999). Calculates Student's t-distribution type statistics, probs and confidence intervals. Also performs diagnostic tests of error terms, such as the Breusch-Godfrey autocorrelation test, the Goldfeld-Quandt heteroscedasticity test, the Conover nonparametric test of homogeneity of variance and the Jarque-Bera normality test

Usage

optimPMG(dLL, maxIter, TetaStart, vecSR, vecLR, dataset, quantity, const)

Arguments

dLL

a parameter indicating the convergence criterion; an optimization algorithm is stopped when an increase in concentrated log-likelihood function (Equation 8 in Pesaran, Shin and Smith (1999)) is less than dLL; the default value is dLL=10^-10

maxIter

a maximum number of iterations; the default value is 200

TetaStart

a vector of first (initial) Teta values, from which the algorithm starts searching for parameters ensuring the maximization of log-likelihood function

vecSR

a list of vectors containing the column numbers of variables in short-run relationships for each group (alternatively a list of vectors containing the variables names instead of column numbers). In each vector of the list the first number must indicate dy (i.e., the dependant variable)

vecLR

a vector containing the column numbers of variables in long-run relationships (alternatively a vector containing the variables names instead of column numbers). The first number must indicate ly (i.e., the lagged dependant variable)

dataset

a panel data set in the form of stacked time series, containing variables of long-run and short-run relationships (i.e., including differentiated and lagged variables)

quantity

a vector of the number of time series observations in each group; in practice, it takes the form c(T1,...,Tn) since the PMG allows the numbers of time series observations to differ freely across groups (if the number of time series observations in each group is the same, then c(T,...,T) and T=T1=T2=...=Tn

const

logical. If TRUE (the default value), the intercept term is added to the model (i.e., to the short-run relationship)

Details

Estimates parameters of long-run and short-run relationships. Also estimates the information matrix as well as standard errors of estimations, as indicated in Equation 13 (Pesaran, Shin and Smith, 1999). Calculates Student's t-distribution type statistics, probs and confidence intervals. Also performs diagnostic tests of error terms, such as the Breusch-Godfrey autocorrelation test, the Goldfeld-Quandt heteroscedasticity test and the Conover nonparametric test of homogeneity of variance and the Jarque-Bera normality test

Value

$LogL

the concentrated log-likelihood function

$dLogL

the incresase of concentrated log-likelihood function in last iteration

$i

the number of iterations performed to achieve convergence

$LR

the estimated parameters of long-run relationships

$SR

the estimated parameters of short-run relationships

$DiagTests

results of diagnostic tests

$residuals

residuals

Author(s)

Lech Kujawski, Piotr Zientara

References

Pesaran, Shin and Smith (1999) <doi:10.1080/01621459.1999.10474156>

Examples

# first import DataExp, i=1...9, T1=T2=...T9=35
data(DataExp)
DataExp[1:5,]
# then prepare lags and diffs using LagPanel and DiffPanel
y10=data.frame(y10=DataExp[,1], row.names=row.names(DataExp))
cpi=data.frame(cpi=DataExp[,7], row.names=row.names(DataExp))
dy10=DiffPanel(variable=y10, quantity=rep(35,9))
dopeness=DiffPanel(variable=DataExp[,6], quantity=rep(35,9))
ly10=LagPanel(variable=y10, quantity=rep(35,9))
diip=DiffPanel(variable=DataExp[,11], quantity=rep(35,9))
dcrisk=DiffPanel(variable=DataExp[,9], quantity=rep(35,9))
ldcrisk=LagPanel(variable=dcrisk, quantity=rep(35,9))
dcpi=DiffPanel(variable=DataExp[,7], quantity=rep(35,9))
ddcpi=DiffPanel(variable=dcpi, quantity=rep(35,9))
ldebt=LagPanel(variable=DataExp[,4], quantity=rep(35,9))
# create homogenous preliminary dataset (containing "NA" as a result of DiffPanel, LagPanel)
dataPanel=cbind(y10, dy10, ly10, DataExp[,6], dopeness, diip,
DataExp[,11], ldcrisk, DataExp[,9], ddcpi, DataExp[,7])
dataPanel=data.frame(dataPanel)
names(dataPanel)=c("y10", "dy10", "ly10", "openess", "dopeness", "diip",
"iip", "ldcrisk", "crisk", "ddcpi", "cpi")
dataPanel[1:5,]
# prepare dataset and quantity for PMG or optimPMG functions using PanelNaOmit
dataPanel=PanelNaOmit(dataset=dataPanel, quantity=rep(35,9))
dataPanel$dataset[1:5,]
dataPanel$quantity
# optimPMG
OptimPmgExp=optimPMG(
dLL=10^-10,
maxIter=200,
TetaStart=rep(x=1, times=4), # note that length(TetaStart)=length(vecLR)-1
vecSR=list(SR1=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR2=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR3=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR4=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR5=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR6=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR7=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR8=c("dy10","dopeness","diip","ldcrisk","ddcpi"),
SR9=c("dy10","dopeness","diip","ldcrisk","ddcpi")),
vecLR=c("ly10","openess","iip","crisk","cpi"),
dataset=dataPanel$dataset,
quantity=dataPanel$quantity,
const=TRUE)
OptimPmgExp