| Title: | Dynamic Panel Multiple Threshold Model with Fixed Effects |
|---|---|
| Description: | Compute the fixed effects dynamic panel threshold model suggested by Ramírez-Rondán (2020) <doi:10.1080/07474938.2019.1624401>, and dynamic panel linear model suggested by Hsiao et al. (2002) <doi:10.1016/S0304-4076(01)00143-9>, where maximum likelihood type estimators are used. Multiple threshold estimation based on Markov Chain Monte Carlo (MCMC) is allowed, and model selection of linear model, threshold model and multiple threshold model is also allowed. |
| Authors: | Bai Hujie [aut, cre, cph] |
| Maintainer: | Bai Hujie <[email protected]> |
| License: | GPL (>= 3) |
| Version: | 1.3.7 |
| Built: | 2024-11-18 05:10:23 UTC |
| Source: | https://github.com/hujiebai/dptm |
A simulation data used for examples
datadata
A simulation data used for examples
None
The dynamic panel linear model with fixed effects
DPML( y, y1 = NULL, x = NULL, w = NULL, var_u = NULL, tt, nn, time_trend = FALSE, time_fix_effects = FALSE, restart = FALSE, x1 = NULL, delty0 = NULL, Only_b = FALSE, display = TRUE )DPML( y, y1 = NULL, x = NULL, w = NULL, var_u = NULL, tt, nn, time_trend = FALSE, time_fix_effects = FALSE, restart = FALSE, x1 = NULL, delty0 = NULL, Only_b = FALSE, display = TRUE )
y |
the dependent variable; vector type input. |
y1 |
the lag dependent variable; vector type input; By default, y1 is NULL, and then y1 will be computed by y automatically. |
x |
the independent variable; matrix type input. |
w |
the variance ratio; By default, is NULL; It must be greater than 1. |
var_u |
the option of variance of error term; By default, is NULL; It must be greater than 0; When meet relevant ERROR, please change the var_u. |
tt |
the length of time period. |
nn |
the number of individuals. |
time_trend |
the time trend; By default, it is FALSE. |
time_fix_effects |
the time fixed effects; By default, it is FALSE. |
restart |
the option of iterations; By default, restart is FALSE, if encounters iteration failure, please set restart as TRUE. |
x1 |
the initial values of independent variable; matrix type input. By default, x1 is NULL, and thus x1 will be computed by x automatically. |
delty0 |
the option of delta_y; By default, delty0 is NULL; Please do not change delty0. |
Only_b |
the option of initial equation;By default, Only_b is FALSE, and if Only_b is TRUE, initial delta y will be a constant C. Please see Hsiao (2002) and Ramírez-Rondán (2020) for more details. |
display |
the option of whether to print the messages of estimated results; By default, the display is TRUE. |
A list containing the following components:
ssemin |
the negaive log-likelihood function value |
Coefs |
parameter estimates containing t-values |
pars |
iterated results for all parameters |
duit |
the first-difference form of residuals |
dy0 |
the first-difference form of dependent variable |
xx |
the independent variables and their initial values |
covariance_matrix |
the covariance matrix |
Ses |
the standard errors of coefs |
Zvalues |
the values of the statistic |
ccd |
the number of independent variables |
coefs |
parameter estimates containing their initial valuess |
DPML(): This is a dynamic panel linear model with fixed effects, which
allows time trend term or time fixed effects.
Hujie Bai
Ramírez-Rondán, N. R. (2020). Maximum likelihood estimation of dynamic panel threshold models. Econometric Reviews, 39(3), 260-276.
Hsiao, C., Pesaran, M. H., & Tahmiscioglu, A. K. (2002). Maximum likelihood estimation of fixed effects dynamic panel data models covering short time periods. Journal of econometrics, 109(1), 107-150.
data("data", package = "DPTM") y <- data$data_test_linear$y q <- data$data_test_linear$q x <- as.matrix(data$data_test_linear$x) z <- as.matrix(data$data_test_linear$z) tt <- data$data_test_linear$tt nn <- data$data_test_linear$nn xx <- cbind(x,z) m1 <- DPML(y=y,x=xx,tt=tt,nn=nn) m1$Coefsdata("data", package = "DPTM") y <- data$data_test_linear$y q <- data$data_test_linear$q x <- as.matrix(data$data_test_linear$x) z <- as.matrix(data$data_test_linear$z) tt <- data$data_test_linear$tt nn <- data$data_test_linear$nn xx <- cbind(x,z) m1 <- DPML(y=y,x=xx,tt=tt,nn=nn) m1$Coefs
DPTS This is a dynamic panel threshold model with fixed effects, which allows multiple thresholds, time trend term or time fixed effects.
DPTS( y, y1 = NULL, x = NULL, q, cvs = NULL, time_trend = FALSE, time_fix_effects = FALSE, x1 = NULL, tt, nn, Th = 1, ms = 1000, burnin = 1000, types = "DREAMzs", ADs = FALSE, r0x = NULL, r1x = NULL, NoY = FALSE, restart = FALSE, Only_b = FALSE, w = NULL, var_u = NULL, delty0 = NULL, nCR = 3, autoburnin = TRUE, sro = 0.1, display = TRUE )DPTS( y, y1 = NULL, x = NULL, q, cvs = NULL, time_trend = FALSE, time_fix_effects = FALSE, x1 = NULL, tt, nn, Th = 1, ms = 1000, burnin = 1000, types = "DREAMzs", ADs = FALSE, r0x = NULL, r1x = NULL, NoY = FALSE, restart = FALSE, Only_b = FALSE, w = NULL, var_u = NULL, delty0 = NULL, nCR = 3, autoburnin = TRUE, sro = 0.1, display = TRUE )
y |
the dependent variable; vector type input. |
y1 |
the lag dependent variable; vector type input; By default, y1 is NULL, and then y1 will be computed by y automatically. |
x |
the independent variable; matrix type input. |
q |
the threshold variable; vector type input. |
cvs |
the set of control variables; matrix type input;By default, cvs is NULL. |
time_trend |
the time trend; By default, it is FALSE. |
time_fix_effects |
the time fixed effects; By default, it is FALSE. |
x1 |
the initial values of independent variable; matrix type input. By default, x1 is NULL, and thus x1 will be computed by x automatically. |
tt |
the length of time period. |
nn |
the number of individuals. |
Th |
the number of thresholds. |
ms |
the length of MCMC chains after burn-in. |
burnin |
the length of burn-in. |
types |
the type of MCMC used; More details see BayesianTools::runMCMC. |
ADs |
the options for MCMC; More details see BayesianTools::runMCMC. |
r0x |
the lower bound of thresholds; By default, r0x is NULL, and thus r0x will be computed by q automatically. |
r1x |
the upper bound of thresholds; By default, r0x is NULL, and thus r1x will be computed by q automatically. |
NoY |
the option of threshold effects on the lag dependent variable; By default, NoY is False, and thus there will be threshold effects on y1. |
restart |
the option of iterations; By default, restart is FALSE, if encounters iteration failure, please set restart as TRUE. |
Only_b |
the option of initial equation;By default, Only_b is FALSE, and if Only_b is TRUE, initial delta y will be a constant C.; Please see Hsiao (2002) and Ramírez-Rondán (2020) for more details. |
w |
the variance ratio; By default, is NULL; It must be greater than 1. |
var_u |
the option of variance of error term; By default, is NULL; It must be greater than 0; When meet relevant ERROR, please change the var_u. |
delty0 |
the option of delta_y; By default, delty0 is NULL; Please do not change delty0. |
nCR |
parameter determining the number of cross-over proposals of DREAM MCMC. If nCR = 1 all parameters are updated jointly. |
autoburnin |
a logical flag indicating of the Gelman and Rubin's convergence diagnostic, whether variables in x should be transformed to improve the normality of the distribution. If set to TRUE, a log transform or logit transform, as appropriate, will be applied. |
sro |
the least ratio of sample in regimes. |
display |
the option of whether to print the messages of estimated results; By default, the display is TRUE. |
A list containing the following components:
ssemin |
the negaive log-likelihood function value |
Ths |
a vector of multiple thresholds in order |
Ths_IC |
a matrix of confidence intervals of all thresholds |
Coefs |
parameter estimates containing t-values |
MCMC_Convergence_Diagnostic |
the Gelman and Rubin's convergence diagnostic results of MCMC sample |
model |
a list of results of DMPL |
MCMC |
an object of class mcmcSampler (if one chain is run) or mcmcSamplerList, more details see BayesianTools::runMCMC |
Hujie Bai
Ramírez-Rondán, N. R. (2020). Maximum likelihood estimation of dynamic panel threshold models. Econometric Reviews, 39(3), 260-276.
Hsiao, C., Pesaran, M. H., & Tahmiscioglu, A. K. (2002). Maximum likelihood estimation of fixed effects dynamic panel data models covering short time periods. Journal of econometrics, 109(1), 107-150.
data("data", package = "DPTM") y <- data$data_test$y q <-data$data_test$q x <- as.matrix(data$data_test$x) z <- as.matrix(data$data_test$z) tt <- data$data_test$tt nn <- data$data_test$nn m1 <- DPTS(y=y,q=q,x=x,cvs = z,tt=tt,nn=nn,Th=1,ms = 100,burnin = 100) m1$Ths m1$Ths_IC m1$Coefs m1$MCMC_Convergence_Diagnostic plot(m1$MCMC)data("data", package = "DPTM") y <- data$data_test$y q <-data$data_test$q x <- as.matrix(data$data_test$x) z <- as.matrix(data$data_test$z) tt <- data$data_test$tt nn <- data$data_test$nn m1 <- DPTS(y=y,q=q,x=x,cvs = z,tt=tt,nn=nn,Th=1,ms = 100,burnin = 100) m1$Ths m1$Ths_IC m1$Coefs m1$MCMC_Convergence_Diagnostic plot(m1$MCMC)
Threshold_Test This is a test for the numer of thresholds, and it is noted that when under H0 the number of Thresholds is 0, this test is the so called threshold existence test.
Threshold_Test( y, y1 = NULL, x = NULL, q, cvs = NULL, time_trend = FALSE, time_fix_effects = FALSE, x1 = NULL, tt, nn, Th = 0, ms = 1000, burnin = 1000, types = "DREAMzs", ADs = FALSE, r0x = NULL, r1x = NULL, NoY = FALSE, restart = FALSE, Only_b = FALSE, w = NULL, var_u = NULL, nCR = 3, autoburnin = TRUE, bt = 100, parallel = TRUE, sro = 0.1, display = TRUE )Threshold_Test( y, y1 = NULL, x = NULL, q, cvs = NULL, time_trend = FALSE, time_fix_effects = FALSE, x1 = NULL, tt, nn, Th = 0, ms = 1000, burnin = 1000, types = "DREAMzs", ADs = FALSE, r0x = NULL, r1x = NULL, NoY = FALSE, restart = FALSE, Only_b = FALSE, w = NULL, var_u = NULL, nCR = 3, autoburnin = TRUE, bt = 100, parallel = TRUE, sro = 0.1, display = TRUE )
y |
the dependent variable; vector type input. |
y1 |
the lag dependent variable; vector type input; By default, y1 is NULL, and then y1 will be computed by y automatically. |
x |
the independent variable; matrix type input. |
q |
the threshold variable; vector type input. |
cvs |
the set of control variables; matrix type input;By default, cvs is NULL. |
time_trend |
the time trend; By default, it is FALSE. |
time_fix_effects |
the time fixed effects; By default, it is FALSE. |
x1 |
the initial values of independent variable; matrix type input. By default, x1 is NULL, and thus x1 will be computed by x automatically. |
tt |
the length of time period. |
nn |
the number of individuals. |
Th |
the number of thresholds. |
ms |
the length of MCMC chains after burn-in. |
burnin |
the length of burn-in. |
types |
the type of MCMC used; More details see BayesianTools::runMCMC. |
ADs |
the options for MCMC; More details see BayesianTools::runMCMC. |
r0x |
the lower bound of thresholds; By default, r0x is NULL, and thus r0x will be computed by q automatically. |
r1x |
the upper bound of thresholds; By default, r0x is NULL, and thus r1x will be computed by q automatically. |
NoY |
the option of threshold effects on the lag dependent variable; By default, NoY is False, and thus there will be threshold effects on y1. |
restart |
the option of iterations; By default, restart is FALSE, if encounters iteration failure, please set restart as TRUE. |
Only_b |
the option of initial equation;By default, Only_b is FALSE, and if Only_b is TRUE, initial delta y will be a constant C. Please see Hsiao (2002) and Ramírez-Rondán (2020) for more details. |
w |
the variance ratio; By default, is NULL; It must be greater than 1. |
var_u |
the option of variance of error term; By default, is NULL; It must be greater than 0; When meet relevant ERROR, please change the var_u. |
nCR |
parameter determining the number of cross-over proposals of DREAM MCMC. If nCR = 1 all parameters are updated jointly. |
autoburnin |
a logical flag indicating of the Gelman and Rubin's convergence diagnostic, whether variables in x should be transformed to improve the normality of the distribution. If set to TRUE, a log transform or logit transform, as appropriate, will be applied. |
bt |
the number of bootstrap. |
parallel |
the option of parallel; By default, parallel is FALSE, when parallel is TRUE, this test will run in parallel. |
sro |
the least ratio of sample in regimes. |
display |
the option of whether to print the messages of estimated results; By default, the display is TRUE. |
A list containing the following components:
ps |
the p-value of test |
crit |
the crit value of test |
LR |
the statistic |
LRs |
a vector of statistics in bootstrap |
Hujie Bai
Ramírez-Rondán, N. R. (2020). Maximum likelihood estimation of dynamic panel threshold models. Econometric Reviews, 39(3), 260-276.
Hsiao, C., Pesaran, M. H., & Tahmiscioglu, A. K. (2002). Maximum likelihood estimation of fixed effects dynamic panel data models covering short time periods. Journal of econometrics, 109(1), 107-150.
data("data", package = "DPTM") y <- data$data_test$y q <- data$data_test$q x <- as.matrix(data$data_test$x) z <- as.matrix(data$data_test$z) tt <- data$data_test$tt nn <- data$data_test$nn ### Examples elapsed time > 5s m1 <- Threshold_Test(y=y,x=x,q=q,cvs=z,tt=tt,nn=nn,Th=0,ms = 500,burnin=500, bt=10,parallel=FALSE) m1$psdata("data", package = "DPTM") y <- data$data_test$y q <- data$data_test$q x <- as.matrix(data$data_test$x) z <- as.matrix(data$data_test$z) tt <- data$data_test$tt nn <- data$data_test$nn ### Examples elapsed time > 5s m1 <- Threshold_Test(y=y,x=x,q=q,cvs=z,tt=tt,nn=nn,Th=0,ms = 500,burnin=500, bt=10,parallel=FALSE) m1$ps