Nimble On Dai Model

Testing nimble method on Dai model

(long data series)

Define our mcmc procedure in Nimble

my_mcmc <- function(code, constants, data, inits, n_iter=1e5, thin = 1e2){
  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$run(n_iter)
  samples <- as.data.frame(as.matrix(Cmcmc$mvSamples))
  samples <- samples[,1:(length(inits)-1)]
  gather(samples)
}

Generate the test data:

set.seed(1000)
max_days <- 500
DF <- seq(0, 2000, length=max_days) # schedule for env degredation (increased dilution)
x <- numeric(max_days)   
x[1] <- 1.76e5 # initial density
   
for(day in 1:(max_days-1)){
 x[day+1] <- dai(x[day], DF = DF[day])
}

raw <- data.frame(t = 1:max_days, x = x)

Detrend:

N <- length(raw$x)
raw$t <- 1:N
detrend <- loess(x ~ t, raw)
data <- data.frame(x = detrend$residuals/sqrt(var(detrend$residuals)))
qplot(raw$t, data$x, geom='line')

OU Model

ou <- nimbleCode({
   theta ~ dunif(1e-10, 100.0)
       r ~ dunif(1e-10, 20.0)
   sigma ~ dunif(1e-10, 100)
    x[1] ~ dunif(0, 100)

  for(t in 1:(N-1)){
    mu[t] <- x[t] + r * (theta - x[t]) 
    x[t+1] ~ dnorm(mu[t], sd = sigma) 
  }
})

ou_constants <- list(N = N)
ou_inits <- list(theta = 0, r = 1e-3, sigma = 1)

Run the mcmc

df <- my_mcmc(code=ou, ou_constants, data, ou_inits)

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     r 0.2075146 0.02739447
2 sigma 0.6156283 0.01975407

Traces

ggplot(df) + 
  geom_line(aes(seq_along(value), value)) + 
  facet_wrap(~key, scale='free')

Posteriors

ggplot(df) + 
  geom_density(aes(value)) + 
  facet_wrap(~key, scale='free')

LSN version

A modified version of the LSN model to explicitly model the changing parameter as a hidden variable changing at constant rate

lsn <- nimbleCode({
   theta ~ dunif(-1e2, 1e2)
 sigma_x ~ dunif(1e-10, 1e2)
       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] * (theta - x[i]) 
    x[i+1] ~ dnorm(mu_x[i], sd = sigma_x) 
    y[i+1] <- y[i] + m * t[i] / t[N] 
    
  }
})

constants <- list(N = N, t = raw$t)
inits <- list(theta = 0, m = 0, sigma_x = 1, y = rep(1,N))
df <- my_mcmc(code=lsn, constants, data, inits)

Summary statistics

summarise(group_by(df, key), mean=mean(value), std=sqrt(var(value)))
Source: local data frame [3 x 3]

      key         mean          std
1       m -0.003984249 0.0005859252
2 sigma_x  0.575993873 0.0185972953
3   theta -0.010341240 0.0410488393

Traces

ggplot(df) + 
  geom_line(aes(seq_along(value), value)) + 
  facet_wrap(~key, scale='free')

Posteriors

ggplot(df) + 
    geom_density(aes(value)) + 
    facet_wrap(~key, scale='free')

LSN, stochastic hidden variable

Define and model and run MCMC

lsn <- nimbleCode({
   theta ~ dunif(-1e2, 1e2)
   sigma_x ~ dunif(1e-10, 1e2)
   sigma_y ~ dunif(1e-10, 1e2)
       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] * (theta - x[i])
    x[i+1] ~ dnorm(mu_x[i], sd = sigma_x)
    mu_y[i] <- y[i] + m * t[i] / t[N]
    y[i+1] ~ dnorm(mu_y[i], sd = sigma_y) 
  }
})

constants <- list(N = N, t = raw$t)
inits <- list(theta = 0, m = 0, sigma_x = 1, sigma_y = 1, y = rep(1,N))
df <- my_mcmc(code=lsn, constants, data, inits)

Summary statistics

summarise(group_by(df, key), mean=mean(value), std=sqrt(var(value)))
Source: local data frame [4 x 3]

      key         mean         std
1       m -0.003001092 0.004531153
2 sigma_x  0.563929151 0.020224310
3 sigma_y  0.048817421 0.036018516
4   theta  0.010243795 0.045674221

Traces

ggplot(df) + 
  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 3.2.1 (2015-06-18)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Debian GNU/Linux stretch/sid

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_0.4.2             ggplot2_1.0.1          
[3] tidyr_0.2.0             regimeshifts_0.0.0.9000
[5] nimble_0.4              yaml_2.1.13            
[7] knitr_1.10.5           

loaded via a namespace (and not attached):
 [1] igraph_1.0.1     Rcpp_0.12.0      magrittr_1.5    
 [4] MASS_7.3-43      munsell_0.4.2    colorspace_1.2-6
 [7] R6_2.1.0         stringr_1.0.0    plyr_1.8.3      
[10] tools_3.2.1      parallel_3.2.1   grid_3.2.1      
[13] gtable_0.1.2     DBI_0.3.1        lazyeval_0.1.10 
[16] digest_0.6.8     assertthat_0.1   reshape2_1.4.1  
[19] formatR_1.2      codetools_0.2-14 evaluate_0.7    
[22] labeling_0.3     stringi_0.5-5    scales_0.2.5    
[25] proto_0.3-10