start.test <- function(name, wpp.year = NULL) cat('\n<=== Starting test of', name, if(!is.null(wpp.year)) paste0('(WPP ', wpp.year, ')') else '', '====\n')
test.ok <- function(name) cat('\n==== Test of', name, 'OK.===>\n')
test.load.UNtfr <- function(wpp.year=2008) {
# read the UN TFR input file
test.name <- 'loading UN TFR file'
start.test(test.name, wpp.year)
tfr <- bayesTFR:::read.UNtfr(wpp.year)
stopifnot(length(dim(tfr$data.object$data))==2)
stopifnot(dim(tfr$data.object$data)[1] > 150)
stopifnot(is.element('last.observed', colnames(tfr$data.object$data)))
stopifnot(length(tfr$data.object$replaced) == 0)
stopifnot(length(tfr$data.object$added) == 0)
test.ok(test.name)
}
test.load.UNtfr.and.my.tfr.file <- function() {
# read the UN TFR input file for WPP 2019
test.name <- 'loading UN TFR file and my_tfr_file'
start.test(test.name)
my.tfr.file <- file.path(find.package("bayesTFR"), 'extdata', 'my_tfr_template.txt')
tfr <- bayesTFR:::read.UNtfr(2019, my.tfr.file=my.tfr.file)
stopifnot(length(dim(tfr$data.object$data))==2)
stopifnot(length(dim(tfr$suppl.data.object$data))==2)
stopifnot(dim(tfr$data.object$data)[1] == 249)
stopifnot(dim(tfr$suppl.data.object$data)[1] == 104)
stopifnot(is.element('last.observed', colnames(tfr$data.object$data)))
stopifnot(length(tfr$data.object$replaced) == 1)
stopifnot(length(tfr$data.object$added) == 0)
stopifnot(length(tfr$suppl.data.object$replaced) == 0)
stopifnot(length(tfr$suppl.data.object$added) == 1)
test.ok(test.name)
}
test.load.UNlocations <- function(wpp.year=2012) {
test.name <- 'loading WPP location file'
start.test(test.name, wpp.year)
tfr <- suppressWarnings(bayesTFR:::read.UNtfr(wpp.year)) # if wpp.year=2008, there are warnings about non-existent tfr_supplemental dataset
locs <- bayesTFR:::read.UNlocations(tfr$data.object$data, wpp.year) # this could give warnings if there are duplicates in UNlocations
stopifnot(length(dim(locs$loc_data)) == 2)
stopifnot(all(is.element(c('country_code', 'include_code'), colnames(locs$loc_data))))
stopifnot(dim(locs$loc_data)[1] > 200)
stopifnot(all(is.element(intersect(c(0,1,2), locs$loc_data[,'include_code']), c(0,1,2))))
test.ok(test.name)
test.name <- 'loading WPP location file with my.locations.file'
start.test(test.name, wpp.year)
new.locations <- data.frame(country_code=9999, name='my location', reg_code=9999, area_code=8888, reg_name='my region', area_name='my area', location_type=4)
f <- tempfile()
write.table(new.locations, file=f, sep='\t', row.names=FALSE)
mylocs <- bayesTFR:::read.UNlocations(tfr$data.object$data, wpp.year, my.locations.file=f)
# stopifnot(dim(mylocs$loc_data)[1] == dim(locs$loc_data)[1]+1) # has one country more
# stopifnot(is.element(9999, mylocs$loc_data$country_code))
e <- new.env()
do.call("data", list("UNlocations", package=paste0("wpp", wpp.year), envir=e))
# this for some reason causes a warning "‘match’ requires vector arguments"
#data("UNlocations", package=paste0("wpp", wpp.year), envir=e)
my.locations <- e$UNlocations[1:100,]
my.locations[which(my.locations$location_type == 4)[1],'area_code'] <- 123456
write.table(my.locations, file=f, sep='\t', row.names=FALSE)
mylocs <- bayesTFR:::read.UNlocations(tfr$data.object$data, wpp.year, my.locations.file=f)
stopifnot(is.element(123456, mylocs$loc_data$area_code))
test.ok(test.name)
unlink(f)
}
test.create.tfr.matrix <- function(wpp.year=2008) {
test.name <- 'creating TFR matrix'
start.test(test.name, wpp.year)
tfr <- bayesTFR:::read.UNtfr(wpp.year)
locs <- bayesTFR:::read.UNlocations(tfr$data.object$data, wpp.year)
tfr.and.regions <- bayesTFR:::get.TFRmatrix.and.regions(tfr$data.object$data, locs$loc_data,
present.year=2009)
tfr.matrix <- tfr.and.regions$tfr_matrix
stopifnot(dim(tfr.matrix)[1] == 12)
stopifnot(rownames(tfr.matrix)[12] == '2008')
test.ok(test.name)
}
test.run.mcmc.simulation <- function(compression='None', wpp.year = 2019) {
sim.dir <- tempfile()
# run MCMC
test.name <- 'running Phase II MCMC'
start.test(test.name, wpp.year)
m <- run.tfr.mcmc(iter=5, nr.chains=1, output.dir=sim.dir, start.year=1950,
compression.type=compression, wpp.year = wpp.year)
stopifnot(m$mcmc.list[[1]]$finished.iter == 5)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 5)
stopifnot(bayesTFR:::tfr.set.identical(m, get.tfr.mcmc(sim.dir), include.output.dir=FALSE))
test.ok(test.name)
# continue MCMC
test.name <- 'continuing Phase II MCMC'
start.test(test.name, wpp.year)
m <- continue.tfr.mcmc(iter=5, output.dir=sim.dir)
stopifnot(m$mcmc.list[[1]]$finished.iter == 10)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 10)
stopifnot(!is.element(900, m$meta$regions$country_code)) # 'World' should not be included
test.ok(test.name)
# run MCMC for an aggregation
test.name <- 'running Phase II MCMC for extra areas'
start.test(test.name, wpp.year)
data.dir <- file.path(find.package("bayesTFR"), 'extdata')
m <- run.tfr.mcmc.extra(sim.dir=sim.dir,
my.tfr.file=file.path(data.dir, 'my_tfr_template.txt'), burnin=0)
stopifnot(is.element(900, m$meta$regions$country_code)) # 'World' should be included
test.ok(test.name)
test.name <- 'running Phase III MCMC'
start.test(test.name, wpp.year)
m3 <- run.tfr3.mcmc(sim.dir=sim.dir, iter=20, thin=1, nr.chains=2, compression.type=compression)
stopifnot(m3$mcmc.list[[1]]$finished.iter == 20)
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 40)
stopifnot(bayesTFR:::tfr.set.identical(m3, get.tfr3.mcmc(sim.dir), include.output.dir=FALSE))
test.ok(test.name)
# continue MCMC
test.name <- 'continuing Phase III MCMC'
start.test(test.name, wpp.year)
m3 <- continue.tfr3.mcmc(sim.dir=sim.dir, iter=5)
stopifnot(m3$mcmc.list[[1]]$finished.iter == 25)
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 50)
test.ok(test.name)
# run prediction
test.name <- 'running projections with classic AR(1)'
start.test(test.name, wpp.year)
pred <- tfr.predict(m, burnin=0, use.tfr3=FALSE, rho=NULL, sigmaAR1=NULL, verbose=FALSE)
spred <- summary(pred)
stopifnot(spred$nr.traj == 10)
stopifnot(is.na(pred$thin3))
stopifnot(!is.element(903, pred$mcmc.set$regions$country_code))
test.ok(test.name)
rho <- pred$rho
sigma <- pred$sigmaAR1
mu <- pred$mu
test.name <- 'running projections with BHM for phase III'
start.test(test.name, wpp.year)
pred <- tfr.predict(m, burnin=0, use.tfr3=TRUE, burnin3=5, verbose=FALSE, replace.output=TRUE)
spred <- summary(pred)
stopifnot(spred$nr.traj == 10)
stopifnot(pred$thin3 == 4)
stopifnot(!is.element(903, pred$mcmc.set$regions$country_code))
test.ok(test.name)
npred <- dim(pred$tfr_matrix_reconstructed)[2]
# run MCMC for another aggregation
test.name <- 'running projections on extra areas'
start.test(test.name, wpp.year)
m <- run.tfr.mcmc.extra(sim.dir=sim.dir, countries=903, burnin=0)
# run prediction only for the area 903
pred <- tfr.predict.extra(sim.dir=sim.dir, verbose=FALSE)
stopifnot(is.element(903, pred$mcmc.set$meta$regions$country_code))
stopifnot(dim(pred$tfr_matrix_reconstructed)[2] == npred+1)
stopifnot(!is.null(bayesTFR:::get.trajectories(pred, 903)$trajectories))
stopifnot(pred$use.tfr3)
test.ok(test.name)
test.name <- 'estimating AR(1) parameters'
start.test(test.name, wpp.year)
ar1pars <- get.ar1.parameters(mu, m$meta)
stopifnot(rho==ar1pars$rho)
stopifnot(sigma==ar1pars$sigmaAR1)
test.ok(test.name)
test.name <- 'shifting the median'
start.test(test.name, wpp.year)
projs <- summary(pred, country='Uganda')$projections
tfr.median.shift(sim.dir, country='Uganda', shift=1.5, from=2051, to=2080)
shifted.pred <- get.tfr.prediction(sim.dir)
shifted.projs <- summary(shifted.pred, country='Uganda')$projections
stopifnot(all(projs[8:13,c(1,3:dim(projs)[2])]+1.5 == shifted.projs[8:13,c(1,3:dim(projs)[2])]))
stopifnot(all(projs[c(1:7, 14:17),c(1,3:dim(projs)[2])] == shifted.projs[c(1:7, 14:17),c(1,3:dim(projs)[2])]))
test.ok(test.name)
test.name <- 'resetting the median'
start.test(test.name, wpp.year)
shifted.pred <- tfr.median.shift(sim.dir, country='Uganda', reset = TRUE)
shifted.projs <- summary(shifted.pred, country='Uganda')$projections
stopifnot(all(projs[,c(1,3:dim(projs)[2])] == shifted.projs[,c(1,3:dim(projs)[2])]))
test.ok(test.name)
test.name <- 'setting the median'
start.test(test.name, wpp.year)
expert.values <- c(2.3, 2.4, 2.4)
cobj <- get.country.object('Uganda', m$meta)
shift <- expert.values - pred$quantiles[cobj$index, '0.5',2:4]
mod.pred <- tfr.median.set(sim.dir, country='Uganda', values=expert.values, years=2024)
mod.projs <- summary(mod.pred, country='Uganda')$projections
stopifnot(all(mod.projs[2:4, c(1,3:dim(projs)[2])]==projs[2:4, c(1,3:dim(projs)[2])]+shift))
stopifnot(all(mod.projs[c(1,5:17), c(1,3:dim(projs)[2])]==projs[c(1,5:17), c(1,3:dim(projs)[2])]))
test.ok(test.name)
test.name <- 'shifting medians to WPP'
tfr.shift.prediction.to.wpp(sim.dir)
shifted.pred <- get.tfr.prediction(sim.dir)
shifted.projs <- summary(shifted.pred, country='Niger')$projections
shifted.projs <- data.table::data.table(shifted.projs)[, year := as.integer(rownames(shifted.projs))]
e <- new.env()
data("tfrprojMed", package = paste0("wpp", wpp.year), envir = e)
wpptfr <- data.table::data.table(e$tfrprojMed)
wpptfrl <- data.table::melt(wpptfr, id.vars = c("country_code", "name"), variable.name = "period")
wpptfrl <- wpptfrl[, year := as.integer(substr(period, 1,4)) + 3][name == "Niger"]
dat <- merge(wpptfrl, shifted.projs[, c("year", "50%"), with = FALSE], by = "year")
stopifnot(all.equal(dat$value, dat[, `50%`]))
stopifnot(!is.null(shifted.pred$median.shift))
stopifnot(length(shifted.pred$median.shift) == nrow(get.countries.table(shifted.pred)))
test.ok(test.name)
test.name <- 'resetting all countries'
tfr.median.reset(sim.dir)
new.pred <- get.tfr.prediction(sim.dir)
stopifnot(is.null(new.pred$median.shift))
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
test.thinned.simulation <- function(compression='None', wpp.year = 2019) {
sim.dir <- tempfile()
# run MCMC
test.name <- 'running thinned Phase II MCMC'
start.test(test.name, wpp.year)
m <- run.tfr.mcmc(iter=10, nr.chains=2, output.dir=sim.dir, thin=2, compression.type=compression,
wpp.year = wpp.year)
stopifnot(m$mcmc.list[[1]]$finished.iter == 10)
stopifnot(m$mcmc.list[[1]]$length == 6)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 20)
stopifnot(get.stored.mcmc.length(m$mcmc.list, burnin=4) == 6)
test.ok(test.name)
# continue MCMC
test.name <- 'continuing thinned Phase II MCMC'
start.test(test.name, wpp.year)
m <- continue.tfr.mcmc(iter=10, output.dir=sim.dir)
stopifnot(m$mcmc.list[[1]]$finished.iter == 20)
stopifnot(m$mcmc.list[[1]]$length == 11)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 40)
stopifnot(get.stored.mcmc.length(m$mcmc.list, burnin=4) == 16)
test.ok(test.name)
# run MCMC for an aggregation
test.name <- 'running thinned MCMC for extra areas'
start.test(test.name, wpp.year)
data.dir <- file.path(find.package("bayesTFR"), 'extdata')
m <- run.tfr.mcmc.extra(sim.dir=sim.dir, my.tfr.file=file.path(data.dir, 'my_tfr_template.txt'), burnin=0)
stopifnot(is.element(900, m$meta$regions$country_code)) # 'World' should be included
test.ok(test.name)
test.name <- 'running thinned Phase III MCMC'
start.test(test.name, wpp.year)
m3 <- run.tfr3.mcmc(sim.dir=sim.dir, iter=20, nr.chains=3, thin=4, compression.type=compression)
stopifnot(m3$mcmc.list[[1]]$finished.iter == 20)
stopifnot(m3$mcmc.list[[1]]$length == 6)
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 60)
stopifnot(get.stored.mcmc.length(m3$mcmc.list, burnin=4) == 12)
test.ok(test.name)
# continue MCMC
test.name <- 'continuing thinned Phase III MCMC'
start.test(test.name, wpp.year)
m3 <- continue.tfr3.mcmc(sim.dir=sim.dir, iter=5)
stopifnot(m3$mcmc.list[[1]]$finished.iter == 25)
stopifnot(m3$mcmc.list[[1]]$length == 7)
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 75)
stopifnot(get.stored.mcmc.length(m3$mcmc.list, burnin=4) == 15)
test.ok(test.name)
# run prediction
test.name <- 'running thinned projections'
start.test(test.name, wpp.year)
pred <- tfr.predict(m, burnin=5, burnin3=3, verbose=FALSE)
spred <- summary(pred)
stopifnot(spred$nr.traj == 16) # 2x8
stopifnot(pred$thin3 == 4)
stopifnot(pred$mcmc.set$mcmc.list[[1]]$finished.iter == 16)
stopifnot(length(pred$mcmc.set$mcmc.list) == 1)
npred <- dim(pred$tfr_matrix_reconstructed)[2]
test.ok(test.name)
test.name <- 'running thinned projections on extra areas'
start.test(test.name, wpp.year)
m <- run.tfr.mcmc.extra(sim.dir=sim.dir, countries=903, burnin=5)
# run prediction only for the area 903
pred <- tfr.predict.extra(sim.dir=sim.dir, verbose=FALSE)
stopifnot(dim(pred$tfr_matrix_reconstructed)[2] == npred+1)
test.ok(test.name)
test.name <- 'plotting DL curves with thinned MCMC'
start.test(test.name, wpp.year)
filename <- tempfile()
png(filename=filename)
DLcurve.plot(m, 903, burnin=5, nr.curves=5)
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'plotting TFR trajectories with thinned MCMC'
start.test(test.name, wpp.year)
filename <- tempfile()
png(filename=filename)
tfr.trajectories.plot(pred, 900, nr.traj=5, pi=c(70,65))
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'getting Phase II parameter traces with thinned MCMC'
start.test(test.name, wpp.year)
traces <- get.tfr.parameter.traces(m$mcmc.list, burnin=5,
thinning.index=c(1, 3, 10, 15))
stopifnot(nrow(traces)==4)
test.ok(test.name)
test.name <- 'getting Phase III parameter traces with thinned MCMC'
start.test(test.name, wpp.year)
m3 <- get.tfr3.mcmc(sim.dir)
traces <- get.tfr3.parameter.traces(m3$mcmc.list, burnin=10,
thinning.index=c(1, 3, 5, 10))
stopifnot(nrow(traces)==4)
test.ok(test.name)
test.name <- 'plotting Phase II parameter density'
start.test(test.name, wpp.year)
filename <- tempfile()
png(filename=filename)
tfr.pardensity.plot(m, burnin=10)
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'plotting Phase III parameter density'
start.test(test.name, wpp.year)
filename <- tempfile()
png(filename=filename)
tfr3.pardensity.plot(m3, burnin=10)
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
test.run.mcmc.simulation.auto <- function() {
sim.dir <- tempfile()
# run MCMC
test.name <- 'running auto Phase II MCMC'
start.test(test.name)
m <- run.tfr.mcmc(iter='auto', output.dir=sim.dir,
auto.conf=list(iter=10, iter.incr=5, max.loops=3, nr.chains=2, thin=1, burnin=5))
stopifnot(get.total.iterations(m$mcmc.list, 0) == 40)
test.ok(test.name)
test.name <- 'continuing auto Phase II MCMC'
start.test(test.name)
m <- continue.tfr.mcmc(iter='auto', output.dir=sim.dir, auto.conf=list(max.loops=2))
stopifnot(get.total.iterations(m$mcmc.list, 0) == 60)
test.ok(test.name)
test.name <- 'running auto Phase III MCMC'
start.test(test.name)
m3 <- run.tfr3.mcmc(sim.dir=sim.dir, iter='auto', thin=1,
auto.conf=list(iter=10, iter.incr=5, max.loops=3, nr.chains=2, thin=1, burnin=5))
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 40)
test.ok(test.name)
test.name <- 'continuing auto Phase III MCMC'
start.test(test.name)
m3 <- continue.tfr3.mcmc(sim.dir=sim.dir, iter='auto', auto.conf=list(max.loops=2))
stopifnot(get.total.iterations(m$mcmc.list, 0) == 60)
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
test.run.mcmc.simulation.auto.parallel <- function() {
sim.dir <- tempfile()
# run MCMC
test.name <- 'running auto Phase II MCMC in parallel'
start.test(test.name)
m <- run.tfr.mcmc(iter='auto', output.dir=sim.dir, parallel=TRUE, cltype='SOCK',
auto.conf=list(iter=10, iter.incr=5, max.loops=3, nr.chains=2, thin=1, burnin=5))
stopifnot(get.total.iterations(m$mcmc.list, 0) == 40)
test.ok(test.name)
test.name <- 'continuing auto Phase II MCMC in parallel'
start.test(test.name)
m <- continue.tfr.mcmc(iter='auto', output.dir=sim.dir, auto.conf=list(max.loops=2), parallel=TRUE, cltype='SOCK')
stopifnot(get.total.iterations(m$mcmc.list, 0) == 60)
test.ok(test.name)
test.name <- 'running auto Phase III MCMC in parallel'
start.test(test.name)
m3 <- run.tfr3.mcmc(sim.dir=sim.dir, iter='auto', thin=1, parallel=TRUE, cltype='SOCK',
auto.conf=list(iter=10, iter.incr=5, max.loops=3, nr.chains=2, thin=1, burnin=5))
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 40)
test.ok(test.name)
test.name <- 'continuing auto Phase III MCMC in parallel'
start.test(test.name)
m3 <- continue.tfr3.mcmc(sim.dir=sim.dir, iter='auto', auto.conf=list(max.loops=2), parallel=TRUE, cltype='SOCK' )
stopifnot(get.total.iterations(m$mcmc.list, 0) == 60)
test.ok(test.name)
test.name <- 'running auto Phase II MCMC in parallel with ClusterOptions'
start.test(test.name)
#snow::setDefaultClusterOptions(type='SOCK')
m <- run.tfr.mcmc(iter='auto', output.dir=sim.dir, parallel=TRUE, replace.output=TRUE,
auto.conf=list(iter=10, iter.incr=5, max.loops=3, nr.chains=2, thin=1, burnin=5))
stopifnot(get.total.iterations(m$mcmc.list, 0) == 40)
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
test.imputation <- function() {
sim.dir <- tempfile()
# run MCMC
test.name <- 'running MCMC with missing values'
start.test(test.name)
my.tfr.file <- file.path(find.package('bayesTFR'), 'extdata', 'UN2010_last_obs.txt')
m <- run.tfr.mcmc(iter=5, nr.chains=1, output.dir=sim.dir, my.tfr.file=my.tfr.file, wpp.year=2010, present.year=2010)
stopifnot(m$mcmc.list[[1]]$finished.iter == 5)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 5)
stopifnot(bayesTFR:::tfr.set.identical(m, get.tfr.mcmc(sim.dir), include.output.dir=FALSE))
# some countries are not DL because of the missing data
# This was the case in UN2008 but not in UN2010
# stopifnot(length(get.countries.index(m$meta)) != get.nr.countries(m$meta))
m3 <- run.tfr3.mcmc(sim.dir=sim.dir, iter=10, nr.chains=1, thin=1, my.tfr.file=my.tfr.file)
stopifnot(m3$mcmc.list[[1]]$finished.iter == 10)
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 10)
stopifnot(bayesTFR:::tfr.set.identical(m3, get.tfr3.mcmc(sim.dir), include.output.dir=FALSE))
test.ok(test.name)
test.name <- 'running projections with imputation'
start.test(test.name)
pred <- tfr.predict(m, burnin=0, burnin3=3, verbose=FALSE)
spred <- summary(pred)
stopifnot(spred$nr.traj == 5)
test.ok(test.name)
test.name <- 'plotting imputed TFR trajectories'
start.test(test.name)
filename <- tempfile()
png(filename=filename)
tfr.trajectories.plot(pred, 'Eritrea', pi=c(90, 54))
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'creating parameter maps with imputation'
filename <- tempfile()
tfr.map(pred, device='png', par.name='Triangle_c4', device.args=list(filename=filename))
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
test.existing.simulation <- function() {
test.name <- 'retrieving Phase II MCMC results'
start.test(test.name)
sim.dir <- file.path(find.package("bayesTFR"), "ex-data", 'bayesTFR.output')
m <- get.tfr.mcmc(sim.dir, low.memory=FALSE, burnin=25, chain.ids=1)
stopifnot(length(m$mcmc.list)==1)
stopifnot(dim(m$mcmc.list[[1]]$traces)[1]==35)
test.ok(test.name)
}
test.DLcurve <- function() {
test.name <- 'plotting DL curves'
start.test(test.name)
sim.dir <- file.path(find.package("bayesTFR"), "ex-data", 'bayesTFR.output')
m <- get.tfr.mcmc(sim.dir)
filename <- tempfile()
png(filename=filename)
DLcurve.plot(m, 'Burkina Faso')
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'obtaining DL curves and DL sigma'
start.test(test.name)
tfr <- seq(1, 8, length=100)
dl <- tfr.country.dlcurves(tfr, m, "Nigeria", burnin=10)
stopifnot(all(dim(dl)==c(50,100)))
dls <- tfr.country.dlcurves(tfr, m, "Nigeria", burnin=0, return.sigma=TRUE)
stopifnot(all(dim(dls$sigma)==c(60,100)))
# world distribution
dlw <- tfr.world.dlcurves(tfr, m, countryUc="Nigeria")
stopifnot(all(dim(dlw)==c(60,100)))
# median of the world DL in the TFR range of 3-4 is smaller than the country-specific median in that range
# visual check:
# DLcurve.plot(m, 'Nigeria')
# lines(tfr, apply(dlw, 2, median))
stopifnot(all(apply(dlw[, tfr > 3 & tfr < 4], 2, median) < apply(dls$dl[, tfr > 3 & tfr < 4], 2, median)))
test.ok(test.name)
}
test.TFRtrajectories <- function() {
test.name <- 'plotting TFR trajectories'
start.test(test.name)
sim.dir <- file.path(find.package("bayesTFR"), "ex-data", 'bayesTFR.output')
pred <- get.tfr.prediction(sim.dir=sim.dir)
filename <- tempfile()
png(filename=filename)
tfr.trajectories.plot(pred, 'Australia')
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'tabulating TFR trajectories'
start.test(test.name)
t <- tfr.trajectories.table(pred, 'Australia', pi=c(90, 80, 70))
stopifnot(all(dim(t) == c(30, 9)))
test.ok(test.name)
}
test.plot.all <- function() {
test.name <- 'plotting TFR trajectories and DL curves for all countries'
start.test(test.name)
sim.dir <- file.path(find.package("bayesTFR"), "ex-data", 'bayesTFR.output')
pred <- get.tfr.prediction(sim.dir=sim.dir)
mc <- get.tfr.mcmc(sim.dir)
dir <- tempfile()
tfr.trajectories.plot.all(pred, output.dir=dir, main='XXX trajs')
trajf <- length(list.files(dir, pattern='png$', full.names=FALSE))
DLcurve.plot.all(mc, output.dir=dir, main='DL XXX', output.type='jpeg')
dlf <- length(list.files(dir, pattern='jpeg$', full.names=FALSE))
unlink(dir, recursive=TRUE)
stopifnot(trajf == get.nr.countries(mc$meta))
stopifnot(dlf == length(mc$meta$id_DL))
test.ok(test.name)
}
test.plot.density <- function() {
test.name <- 'plotting parameter density'
start.test(test.name)
sim.dir <- file.path(find.package("bayesTFR"), "ex-data", 'bayesTFR.output')
pred <- get.tfr.prediction(sim.dir=sim.dir)
filename <- tempfile()
png(filename=filename)
tfr.pardensity.cs.plot('Ireland', pred)
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
}
test.plot.map <- function() {
test.name <- 'creating TFR maps via rworldmap'
start.test(test.name)
sim.dir <- file.path(find.package("bayesTFR"), "ex-data", 'bayesTFR.output')
pred <- get.tfr.prediction(sim.dir=sim.dir)
filename <- tempfile()
tfr.map(pred, year=2043, device='png', device.args=list(filename=filename))
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'creating parameter maps'
filename <- tempfile()
tfr.map(pred, year=1985, device='png', par.name='gamma_2', device.args=list(filename=filename))
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'creating lambda maps'
filename <- tempfile()
tfr.map(pred, device='png', par.name='lambda', quantile=0.7,
device.args=list(filename=filename), numCats=20, catMethod="pretty")
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'creating TFR maps using ggplot2'
filename <- paste0(tempfile(), ".png")
tfr.ggmap(pred, file.name = filename)
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
tfr.ggmap(pred, same.scale = TRUE, file.name = filename)
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
tfr.ggmap(pred, same.scale = TRUE, year = 2100, file.name = filename)
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
}
test.get.parameter.traces <- function() {
test.name <- 'getting parameter traces'
start.test(test.name)
sim.dir <- file.path(find.package("bayesTFR"), "ex-data", 'bayesTFR.output')
m <- get.tfr.mcmc(sim.dir, low.memory=TRUE)
traces <- get.tfr.parameter.traces(m$mcmc.list, burnin=15,
thinning.index=c(4, 25, 29))
stopifnot(nrow(traces)==3)
m.check <- get.tfr.mcmc(sim.dir, low.memory=FALSE, burnin=15)
stopifnot(traces[1,'chi']==m.check$mcmc.list[[1]]$traces[4,'chi'])
#stopifnot(all(traces[c(2,3),'chi']==m.check$mcmc.list[[2]]$traces[c(10,14),'chi'])) # in case of two chains
stopifnot(all(traces[c(2,3),'chi']==m.check$mcmc.list[[1]]$traces[c(25,29),'chi']))
m.check <- get.tfr.mcmc(sim.dir, low.memory=FALSE, burnin=30)
traces <- get.tfr.parameter.traces(m$mcmc.list, burnin=30, thin=8)
stopifnot(nrow(traces)==4)
stopifnot(traces[2,'psi']==m.check$mcmc.list[[1]]$traces[9,'psi'])
#stopifnot(traces[4,'psi']==m.check$mcmc.list[[2]]$traces[10,'psi'])
stopifnot(traces[4,'psi']==m.check$mcmc.list[[1]]$traces[25,'psi'])
test.ok(test.name)
}
test.median.adjust <- function() {
test.name <- 'adjusting median'
start.test(test.name)
sim.dir <- file.path(find.package("bayesTFR"), "ex-data", 'bayesTFR.output')
m <- get.tfr.mcmc(sim.dir)
pred.dir <- tempfile()
pred <- tfr.predict(m, burnin=0, output.dir=pred.dir, nr.traj=10, save.as.ascii=0, verbose=FALSE, use.tfr3=FALSE)
country.obj1 <- get.country.object('Kenya', m$meta)
country.obj2 <- get.country.object('Mongolia', m$meta)
new.pred <- tfr.median.adjust(pred.dir, countries=c('Kenya', 'Mongolia'))
orig.median1 <- get.median.from.prediction(new.pred, country.obj1$index, country.obj1$code, adjusted=FALSE)
new.median1 <- get.median.from.prediction(new.pred, country.obj1$index, country.obj1$code, adjusted=TRUE)
orig.median2 <- get.median.from.prediction(new.pred, country.obj2$index, country.obj2$code, adjusted=FALSE)
new.median2 <- get.median.from.prediction(new.pred, country.obj2$index, country.obj2$code, adjusted=TRUE)
stopifnot(sum(orig.median1==new.median1)==1) # only the present year should be equal
stopifnot(sum(orig.median2==new.median2)==1)
test.ok(test.name)
test.name <- 'plotting adjusted TFR trajectories'
start.test(test.name)
filename <- tempfile()
png(filename=filename)
tfr.trajectories.plot(new.pred, 'Kenya', adjusted.only=FALSE)
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'writing summary files'
start.test(test.name)
write.projection.summary(sim.dir, adjusted=FALSE, output.dir=pred.dir)
write.projection.summary(sim.dir, adjusted=TRUE, output.dir=pred.dir)
test.ok(test.name)
unlink(pred.dir)
}
test.estimate.mcmc.with.suppl.data <- function() {
sim.dir <- tempfile()
# run MCMC
test.name <- 'estimating MCMC using supplemental data'
start.test(test.name)
m <- run.tfr.mcmc(nr.chains=1, iter=30, thin=1, output.dir=sim.dir, start.year=1750, seed=1)
stopifnot(length(m$meta$suppl.data$regions$country_code) == 103)
stopifnot(all(dim(m$meta$suppl.data$tfr_matrix) == c(40, 103)))
test.ok(test.name)
# continue MCMC
test.name <- 'continuing MCMC with supplemental data'
start.test(test.name)
m <- continue.tfr.mcmc(iter=10, output.dir=sim.dir)
stopifnot(m$mcmc.list[[1]]$finished.iter == 40)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 40)
stopifnot(!is.element(900, m$meta$regions$country_code)) # 'World' should not be included
test.ok(test.name)
# run MCMC for an aggregation
test.name <- 'estimating MCMC for extra areas with supplemental data'
start.test(test.name)
data.dir <- file.path(find.package("bayesTFR"), 'extdata')
m <- run.tfr.mcmc.extra(sim.dir=sim.dir,
my.tfr.file=file.path(data.dir, 'my_tfr_template.txt'), burnin=0, verbose=TRUE)
stopifnot(is.element(900, m$meta$regions$country_code)) # 'World' should be included
stopifnot(is.element(900, m$meta$suppl.data$regions$country_code))
test.ok(test.name)
# run prediction
test.name <- 'running projections for simulation with supplemental data'
start.test(test.name)
pred <- tfr.predict(m, burnin=10, verbose=FALSE, save.as.ascii=0, rho=NULL, sigmaAR1=NULL, use.tfr3=FALSE)
spred <- summary(pred)
stopifnot(spred$nr.traj == 30)
stopifnot(!is.element(903, pred$mcmc.set$regions$country_code))
npred <- dim(pred$tfr_matrix_reconstructed)[2]
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
test.subnational.predictions <- function() {
sim.dir <- tempfile()
test.name <- 'predicting subnational TFR'
start.test(test.name)
my.subtfr.file <- file.path(find.package("bayesTFR"), 'extdata', 'subnational_tfr_template.txt')
nat.dir <- file.path(find.package("bayesTFR"), "ex-data", "bayesTFR.output")
# Subnational projections for Australia and Canada; start 1 time period before national projections
preds <- tfr.predict.subnat(c(36, 124), my.tfr.file=my.subtfr.file,
sim.dir=nat.dir, output.dir=sim.dir, start.year=2011)
stopifnot(all(names(preds) %in% c("36", "124")))
stopifnot(nrow(get.countries.table(preds[["36"]]))==8)
filename <- tempfile()
pdf(file=filename)
tfr.trajectories.plot(preds[["36"]], "Queensland")
tfr.trajectories.plot(preds[["124"]], "Quebec")
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
unlink(sim.dir, recursive=TRUE)
# Subnational projections for Canada; start year after the start of national projections
sim.dir <- tempfile()
preds <- tfr.predict.subnat("Canada", my.tfr.file=my.subtfr.file,
sim.dir=nat.dir, output.dir=sim.dir, start.year=2018)
stopifnot(all(names(preds) %in% c("124")))
spred <- summary(preds[["124"]], "Quebec")
stopifnot(min(spred$projection.years) == 2018)
unlink(sim.dir, recursive=TRUE)
# Subnational projections for Canada with one region (Alberta) having imputed values
sim.dir <- tempfile()
mytfr <- read.delim(my.subtfr.file, check.names = FALSE)
mytfr[mytfr$country_code == 124 & mytfr$reg_code == 658, "last.observed"] <- 1999
my.subtfr.file.mis <- tempfile()
write.table(mytfr, file = my.subtfr.file.mis, row.names = FALSE, quote = FALSE, sep = "\t")
preds <- tfr.predict.subnat("Canada", my.tfr.file=my.subtfr.file.mis,
sim.dir=nat.dir, output.dir=sim.dir, start.year=2013)
obs <- preds[["124"]]$mcmc.set$meta$tfr_matrix_observed[,"658"]
recon <- preds[["124"]]$tfr_matrix_reconstructed[,"658"]
stopifnot(all(is.na(obs[c("2003", "2008")]))) # observed is NA
stopifnot(all(!is.na(obs["1998"]))) # last observed non-NA
stopifnot(all(!is.na(recon[c("2003", "2008")]))) # missing values were reconstructed
unlink(sim.dir, recursive=TRUE)
unlink(my.subtfr.file.mis)
# Subnational projections for all countries in the file
sim.dir <- tempfile()
preds <- tfr.predict.subnat(c(36, 218, 124), my.tfr.file=my.subtfr.file, sim.dir=nat.dir,
output.dir=sim.dir, end.year=2050)
stopifnot(length(preds) == 3)
# Retrieve results for all countries
preds <- get.regtfr.prediction(sim.dir)
stopifnot(length(preds) == 3)
t <- tfr.trajectories.table(preds[["218"]], "Bolivar")
stopifnot(all(dim(t) == c(17, 7)))
# Retrieve results for one country
pred <- get.regtfr.prediction(sim.dir, 218)
spred <- summary(pred, "Loja")
stopifnot(max(spred$projection.years) == 2048)
# Retrieve trajectories
trajs <- get.tfr.trajectories(preds[["124"]], "Alberta")
stopifnot(all(dim(trajs) == c(7, 30)))
unlink(sim.dir, recursive=TRUE)
# Annual subnational projections
my.subtfr.file.annual <- tempfile()
dat <- data.table::fread(my.subtfr.file) # load the data and save only the last time period, pretending this is a single year data point
dat <- dat[, .(country_code, country, reg_code, name, `2008` = `2005-2010`)]
data.table::fwrite(dat, file = my.subtfr.file.annual, sep = "\t") # save it
sim.dir <- tempfile()
preds <- tfr.predict.subnat(c(36, 218, 124), my.tfr.file=my.subtfr.file.annual, sim.dir=nat.dir,
output.dir=sim.dir, start.year = 2009, end.year=2030, annual = TRUE)
# Retrieve trajectories
trajs <- get.tfr.trajectories(preds[["124"]], "Alberta")
stopifnot(all(dim(trajs) == c(23, 30))) # from 2008 to 2030
stopifnot(all(c(2021, 2029) %in% as.integer(rownames(trajs))))
pred <- get.regtfr.prediction(sim.dir, 218)
spred <- summary(pred, "Manabi")
stopifnot(max(spred$projection.years) == 2030)
t <- tfr.trajectories.table(preds[["124"]], "Ontario")
stopifnot(max(as.integer(rownames(t))) == 2030)
stopifnot(all(c(2009, 2014, 2029) %in% as.integer(rownames(t))))
unlink(sim.dir, recursive=TRUE)
unlink(my.subtfr.file.annual)
test.ok(test.name)
}
test.reproduce.simulation <- function() {
sim.dir <- tempfile()
test.name <- 'reproducing simulation'
start.test(test.name)
seed <- 1234
m1 <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, seed = seed)
res1 <- summary(m1)$phase2$results$statistics
m2 <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, seed = seed, replace.output = TRUE)
res2 <- summary(m2)$phase2$results$statistics
stopifnot(!is.null(res1) && all(res1 == res2))
m3 <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, replace.output = TRUE) # no seed
res3 <- summary(m3)$phase2$results$statistics
stopifnot(!all(res3 == res2))
# parallel
m1p <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, seed = seed, replace.output = TRUE, parallel = TRUE, ft_verbose = TRUE)
res1p <- summary(m1p)$phase2$results$statistics
m2p <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, seed = seed, replace.output = TRUE, parallel = TRUE, ft_verbose = TRUE)
res2p <- summary(m2p)$phase2$results$statistics
stopifnot(!is.null(res1p) && all(res1p == res2p))
m3p <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, replace.output = TRUE, parallel = TRUE, ft_verbose = TRUE) # no seed
res3p <- summary(m3p)$phase2$results$statistics
stopifnot(!all(res3p == res2p))
# with uncertainty
mu1 <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, seed = seed, uncertainty = TRUE, replace.output = TRUE)
resu1 <- summary(mu1)$phase2$results$statistics
mu2 <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, seed = seed, uncertainty = TRUE, replace.output = TRUE)
resu2 <- summary(mu2)$phase2$results$statistics
mu3 <- run.tfr.mcmc(iter=5, nr.chains=2, output.dir=sim.dir, start.year=1950, uncertainty = TRUE, replace.output = TRUE)
stopifnot(!is.null(resu1) && all(resu1 == resu2)) # includes both phase 2 and 3 parameters
stopifnot(!all(resu1 == summary(mu3)$phase2$results$statistics))
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
test.run.mcmc.simulation.with.uncertainty <- function(wpp.year = 2019) {
sim.dir <- tempfile()
# run MCMC
test.name <- 'running MCMC with uncertainty'
start.test(test.name, wpp.year)
m <- run.tfr.mcmc(iter=5, nr.chains=1, output.dir=sim.dir, uncertainty = TRUE, wpp.year = wpp.year)
stopifnot(m$mcmc.list[[1]]$finished.iter == 5)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 5)
stopifnot(bayesTFR:::tfr.set.identical(m, get.tfr.mcmc(sim.dir), include.output.dir=FALSE))
# continue MCMC
test.name <- 'continuing MCMC with uncertainty'
start.test(test.name, wpp.year)
m <- continue.tfr.mcmc(iter=5, output.dir=sim.dir)
stopifnot(m$mcmc.list[[1]]$finished.iter == 10)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 10)
stopifnot(m$mcmc.list[[1]]$uncertainty)
test.ok(test.name)
# summary of MCMC
test.name <- 'summary of MCMC with uncertainty'
start.test(test.name, wpp.year)
sm <- summary(m, country = "Niger")
sm2 <- summary(m, country = "Japan")
smg <- summary(m)
stopifnot(!sm$phase3$estimation)
stopifnot(sm2$phase3$estimation)
stopifnot(!is.null(smg$phase3))
test.ok(test.name)
# run MCMC for an extra country
test.name <- 'running MCMC with uncertainty for extra country'
start.test(test.name, wpp.year)
countries <- c(124, 840)
m <- run.tfr.mcmc.extra(sim.dir=sim.dir, countries = countries, iso.unbiased = countries, burnin=0, uncertainty = TRUE)
stopifnot(sum(m$meta$regions$country_code %in% countries) == length(countries)) # countries were replaced and not added
test.ok(test.name)
test.name <- 'checking Phase III with uncertainty'
start.test(test.name, wpp.year)
m3 <- get.tfr3.mcmc(sim.dir=sim.dir)
stopifnot(m3$mcmc.list[[1]]$finished.iter == 10)
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 10)
test.ok(test.name)
# run prediction
test.name <- 'running projections with uncertainty'
start.test(test.name, wpp.year)
pred <- tfr.predict(m, burnin=0, burnin3=0, uncertainty = TRUE)
spred <- summary(pred)
stopifnot(spred$nr.traj == 10)
test.ok(test.name)
npred <- dim(pred$tfr_matrix_reconstructed)[2]
# run MCMC and prediction for extra country
test.name <- 'running projections with uncertainty for extra country'
start.test(test.name, wpp.year)
m <- run.tfr.mcmc.extra(sim.dir=sim.dir, countries=840, burnin=0, uncertainty = TRUE)
# run prediction only for 840
pred <- tfr.predict.extra(sim.dir=sim.dir, countries = 840, verbose=FALSE, uncertainty = TRUE)
stopifnot(dim(pred$tfr_matrix_reconstructed)[2] == npred)
stopifnot(!is.null(bayesTFR:::get.trajectories(pred, 840)$trajectories))
stopifnot(pred$use.tfr3)
test.ok(test.name)
test.name <- 'getting bias and sd models'
start.test(test.name, wpp.year)
bias_sd <- tfr.bias.sd(m, "Nigeria")
stopifnot(!is.null(bias_sd))
test.ok(test.name)
test.name <- 'plotting uncertainty'
start.test(test.name, wpp.year)
filename <- tempfile()
png(filename=filename)
g <- tfr.estimation.plot(pred, "US", save.image = FALSE)
print(g)
dev.off()
size <- file.info(filename)['size']
unlink(filename)
stopifnot(size > 0)
test.ok(test.name)
test.name <- 'write projection summary'
start.test(test.name, wpp.year)
write.projection.summary(sim.dir, est.uncertainty = FALSE)
write.projection.summary(sim.dir, est.uncertainty = TRUE)
# TODO: add a check of something
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
test.run.annual.simulation <- function(wpp.year = 2019) {
sim.dir <- tempfile()
test.name <- 'running MCMC with annual data'
start.test(test.name, wpp.year)
m <- run.tfr.mcmc(iter = 5, nr.chains = 1, output.dir = sim.dir, annual = TRUE, ar.phase2 = TRUE,
present.year = 2018, wpp.year = wpp.year)
stopifnot(get.total.iterations(m$mcmc.list, 0) == 5)
stopifnot(all(1953:2018 %in% rownames(m$meta$tfr_matrix)))
stopifnot(bayesTFR:::tfr.set.identical(m, get.tfr.mcmc(sim.dir), include.output.dir=FALSE))
stopifnot("rho_phase2" %in% tfr.parameter.names(meta = m$meta))
test.ok(test.name)
test.name <- 'running annual MCMC for extra country'
start.test(test.name, wpp.year)
countries <- c(900, 908)
m <- run.tfr.mcmc.extra(sim.dir = sim.dir, countries = countries, burnin = 0)
stopifnot(countries %in% m$meta$regions$country_code)
test.ok(test.name)
test.name <- 'running Phase III MCMC with annual data'
start.test(test.name, wpp.year)
m3 <- run.tfr3.mcmc(sim.dir=sim.dir, iter=20, thin=1, nr.chains=2)
stopifnot(get.total.iterations(m3$mcmc.list, 0) == 40)
stopifnot(bayesTFR:::tfr.set.identical(m3, get.tfr3.mcmc(sim.dir), include.output.dir=FALSE))
test.ok(test.name)
test.name <- 'running annual projections'
start.test(test.name, wpp.year)
pred <- tfr.predict(m, burnin=0, burnin3=0)
spred <- summary(pred)
stopifnot(spred$nr.traj == 5)
stopifnot(all(2019:2100 %in% spred$projection.years))
stopifnot(all(c(908, 900) %in% get.countries.table(pred)$code))
test.ok(test.name)
unlink(sim.dir, recursive=TRUE)
}
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.