BridgeFixedPanel: Sparse Change Point Model with Fixed Effect

In jongheepark/BridgeChange: Bayesian Change-point models with Bridge prior for analyzing time-series and panel data.

Sparse Change Point Model with Fixed Effect

Description

Sparse Change Point Model with Fixed Effect

Usage

BridgeFixedPanel(
  formula,
  data,
  index,
  model,
  effect,
  standardize = TRUE,
  interaction = 1,
  a = NULL,
  b = NULL,
  n.break = 1,
  alpha.MH = FALSE,
  mcmc = 100,
  burn = 100,
  verbose = 100,
  thin = 1,
  c0 = 0.1,
  d0 = 0.1,
  r0 = 1,
  R0 = 1,
  nu.shape = 2,
  nu.rate = 2,
  alpha = 1,
  Waic = FALSE,
  marginal = FALSE
)

Arguments

data	Data.frame object.
index	String vector for unit and time index variables. For example, index = c("unit", "year").
model	Model (c("within","between", "pooling")).
effect	Effect (c("individual", "time", "twoways")).
interaction	If interaction = 1, no interaciton. If interaction = 2, only two-way interaciton. Interaction can be up to K, which is the rank of the model matrix.
a	a is the shape1 beta prior for transition probabilities. By default, the expected duration is computed and corresponding a and b values are assigned. The expected duration is the sample period divided by the number of states.
b	b is the shape2 beta prior for transition probabilities. By default, the expected duration is computed and corresponding a and b values are assigned. The expected duration is the sample period divided by the number of states.
n.break	Number of breaks. If n.break = 0, it simply runs fixed effect model with shrinkage prior on coefficients.
alpha.MH	If TRUE, alpha is updated by MH algorithm. By default, it is updated by griddy gibbs.
mcmc	The number of MCMC iterations after burn-in.
verbose	A switch which determines whether or not the progress of the sampler is printed to the screen. If verbose is greater than 0, the iteration number and the posterior density samples are printed to the screen every verboseth iteration.
thin	The thinning interval used in the simulation. The number of MCMC iterations must be divisible by this value.
c0	Hyperparam
d0	= 0.1
nu.shape	=2.0
nu.rate	=2.0
alpha	= 1
fomula	Inherited from lm. For example, Y ~ X + Z.
burnin	The number of burn-in iterations for the sampler.

Author(s)

Jong Hee Park, and Soichiro Yamauchi [e-mail]

Examples

####################################
## Alvarez et al. data
####################################
library("pcse")
set.seed(1999)
data("agl")
data = agl
model = "within"
index = c('country', 'year')
effect = 'time'
formula <- growth ~ lagg1 + opengdp + openex + openimp + leftc + central + inter



#########################
## inspection of residuals
#########################
pdata   <- pdata.frame(data, index)
pm <- plm(formula, data = pdata, model = model, effect = effect)
                                        # 
## plot of panel residuals 1
coplot(pm$residuals ~ pdata[,index[2]]|pdata[,index[1]], data=pdata, ## number=length(unique(pdata[,index[1]])),
       overlap=.1, col="brown", type="l", panel = panel.smooth, xlab="panel residuals by group and time")

## plot of panel residuals 2
coplot(pm$residuals ~ pdata[,index[2]]|pdata[,index[1]], type="b", data=pdata)


#########################
## HMBB estimation
#########################
## Be aware that 100 runs are too short for analysis. 
mcmc = 100; burn = 100; verbose = 100; thin = 1;
formula <- growth ~ lagg1 + opengdp + openex + openimp + leftc + central + inter
agl.cp0 <- BridgeFixedPanel(formula=formula, data = data, model = model, index = index, effect = effect,
                            mcmc=mcmc, verbose=verbose, Waic = TRUE, 
                            n.break = 0)
agl.cp1 <- BridgeFixedPanel(formula=formula, data = data, model = model, index = index, effect = effect,
                            mcmc=mcmc, verbose=verbose, Waic = TRUE, 
                            n.break = 1)
agl.cp2 <- BridgeFixedPanel(formula=formula, data = data, model = model, index = index, effect = effect,
                            mcmc=mcmc, verbose=verbose, Waic = TRUE, 
                            n.break = 2)

## model selection by WAIC
waic <- WaicCompare(list(agl.cp0, agl.cp1, agl.cp2), print = TRUE)
plotWaic(waic)

par(mfrow=c(1, 2))
MCMCpack::plotState(agl.cp1, start=1970, legend.control =c(1970, 0.85), main="One break")
MCMCpack::plotState(agl.cp2, start=1970, legend.control =c(1970, 0.85), main="Two breaks")

####################################
## dotplot over time
## time-varying movements of selected covariates
####################################
## all covariates
dotplotRegime(agl.cp1, hybrid=FALSE, start = 1970, location.bar=12, x.location="default",
              text.cex=0.8, main="Time-varying Movements of All Covariates")
## label as a legend
dotplotRegime(agl.cp1, hybrid=FALSE, start = 1970, location.bar=12, x.location="legend",
              text.cex=0.8, main="Time-varying Movements of All Covariates")

## leftc only
## select works like grep()
dotplotRegime(agl.cp1, hybrid=FALSE, start = 1970, location.bar=12, x.location="static",
              text.cex=0.8, select="left", main=("Time-varying Movements of Left Party"))

jongheepark/BridgeChange documentation built on Jan. 12, 2023, 4:48 p.m.