A simpler Nimble model
library("nimble")
library("regimeshifts")
library("tidyr")
library("ggplot2")
library("dplyr")
LSN version
A further simplified version:
- fix
sigma_x
to unity andtheta
to zero, reflecting the detrending and scaling. - constrain
sigma_y
with a tight prior
code <- nimbleCode({
# sigma_x ~ dunif(1e-10, 1e2)
## highly constrained
sigma_y ~ dgamma(10, 1000)
## Uninformative
m ~ dunif(-1e2, 1e2)
x[1] ~ dunif(-1e3, 1e3)
y[1] ~ dunif(-1e3, 1e3)
for(i in 1:(N-1)){
mu_x[i] <- x[i] - y[i] * x[i]
x[i+1] ~ dnorm(mu_x[i], sd = 1)
y[i+1] ~ dnorm(y[i] + m * t[i] / t[N], sd = sigma_y)
}
})
Generate the test data:
set.seed(1000)
N <- 50
DF <- seq(0, 2000, length=N) # schedule for env degredation (increased dilution)
x <- numeric(N)
x[1] <- 1.76e5 # initial density
for(day in 1:(N-1)){
x[day+1] <- dai(x[day], DF = DF[day])
}
raw <- data.frame(t = 1:N, x = x)
Detrend:
raw$t <- 1:N
detrend <- loess(x ~ t, raw)
sigma <- sqrt(var(detrend$residuals))
data <- data.frame(x = detrend$residuals/sigma)
qplot(raw$t, data$x, geom='line')
constants <- list(N = N, t = raw$t)
inits <- list(m = 0, sigma_y = .01, y = rep(1,N))
thin <- 1e2
n_iter <- 1e6
Rmodel <- nimbleModel(code = code, constants = constants, data = data, inits = inits)
Cmodel <- compileNimble(Rmodel)
mcmcspec <- configureMCMC(Rmodel, print=FALSE, thin=thin)
Rmcmc <- buildMCMC(mcmcspec)
Cmcmc <- compileNimble(Rmcmc, project = Cmodel)
Cmcmc$run(n_iter)
NULL
samples <- as.data.frame(as.matrix(Cmcmc$mvSamples))
samples <- samples[,1:(length(inits) - 1)]
df <- gather(samples)
Summary statistics
summarise(group_by(df, key), mean=mean(value), std=sqrt(var(value)))
Source: local data frame [2 x 3]
key mean std
1 m -0.02118436 0.02327253
2 sigma_y 0.01001409 0.00316794
Traces
ggplot(sample_n(df, 2e2)) +
geom_line(aes(seq_along(value), value)) +
facet_wrap(~key, scale='free')
Posteriors
ggplot(df) +
geom_density(aes(value)) +
facet_wrap(~key, scale='free')
Block sampler
constants <- list(N = N, t = raw$t)
inits <- list(m = 0, sigma_y = .01, y = rep(1,N))
Rmodel <- nimbleModel(code = code, constants = constants, data = data, inits = inits)
Cmodel <- compileNimble(Rmodel)
mcmcspec <- configureMCMC(Rmodel, print=FALSE,thin=thin)
mcmcspec$addSampler("RW_block", list(targetNodes=c('m','sigma_y'), adaptInterval=100))
[53] RW_block sampler; targetNodes: m, sigma_y, adaptive: TRUE, adaptScaleOnly: FALSE, adaptInterval: 100, scale: 1, propCov: identity
Rmcmc <- buildMCMC(mcmcspec)
Cmcmc <- compileNimble(Rmcmc, project = Cmodel)
Cmcmc$run(n_iter)
NULL
samples <- as.data.frame(as.matrix(Cmcmc$mvSamples))
samples <- samples[,1:(length(inits)-1)]
df <- gather(samples)
df <- my_mcmc(code=lsn, constants, data, inits)
Error in eval(expr, envir, enclos): could not find function "my_mcmc"
Summary statistics
summarise(group_by(df, key), mean=mean(value), std=sqrt(var(value)))
Source: local data frame [2 x 3]
key mean std
1 m -0.025367634 0.027055911
2 sigma_y 0.009965489 0.003102902
Traces
ggplot(sample_n(df, 1e3)) +
geom_line(aes(seq_along(value), value)) +
facet_wrap(~key, scale='free')
Posteriors
ggplot(df) +
geom_density(aes(value)) +
facet_wrap(~key, scale='free')
sessionInfo()
R version 3.1.3 RC (2015-03-06 r67947)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Debian GNU/Linux 8 (jessie)
locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
[3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
[5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=en_US.UTF-8 LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
attached base packages:
[1] methods stats graphics grDevices utils datasets
[7] base
other attached packages:
[1] dplyr_0.4.1 ggplot2_1.0.0
[3] tidyr_0.2.0 regimeshifts_0.0.0.9000
[5] nimble_0.3-1 yaml_2.1.13
[7] knitr_1.9
loaded via a namespace (and not attached):
[1] assertthat_0.1 codetools_0.2-11 colorspace_1.2-6
[4] DBI_0.3.1 digest_0.6.8 evaluate_0.5.5
[7] formatR_1.0 grid_3.1.3 gtable_0.1.2
[10] igraph_0.7.1 labeling_0.3 lazyeval_0.1.10
[13] magrittr_1.5 MASS_7.3-39 munsell_0.4.2
[16] parallel_3.1.3 plyr_1.8.1 proto_0.3-10
[19] Rcpp_0.11.5 reshape2_1.4.1 scales_0.2.4
[22] stringr_0.6.2 tools_3.1.3