In jtipton25/pgR: Bayesian Polya-Gamma Regression
We are testing the spatial Polya-gamma linear model pg_splm()
knitr::opts_chunk$set(fig.width = 16, fig.height = 9)
library(pgR)
library(mvnfast)
# library(MCMCpack)
library(splines)
library(tidyverse)
library(patchwork)
simulate some data
set.seed(11)
N <- 30^2
J <- 6
p <- 2
## setup the spatial process
locs <- expand.grid(
seq(0, 1, length = sqrt(N)),
seq(0, 1, length = sqrt(N))
)
D <- fields::rdist(locs)
## use the same parameters for each of the J-1 components
tau2 <- 0.4
theta <- log(0.1)
Sigma <- tau2 * correlation_function(D, theta, corr_fun = "exponential")
psi <- t(rmvn(J-1, rep(0, N), Sigma))
## setup the fixed effects process
X <- cbind(1, matrix(runif(N*p), N, p))
beta <- matrix(rnorm((J-1) * ncol(X), 0, 0.25), ncol(X), (J-1))
## make the intercepts smaller to reduce stochastic ordering effect
beta[1, ] <- beta[1, ] - seq(from = 2, to = 0, length.out = J-1)
eta <- X %*% beta + psi
pi <- eta_to_pi(eta)
Y <- matrix(0, N, J)
for (i in 1:N) {
Y[i, ] <- rmultinom(1, 500, pi[i, ])
}
Y_prop <- counts_to_proportions(Y)
Plot the simulated data
## put data into data.frame for plotting
dat <- data.frame(
Y = c(Y_prop),
lon = locs[, 1],
lat = locs[, 2],
species = factor(rep(1:J, each = N)),
pi = c(pi)
)
p_simulated <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) +
geom_raster() +
scale_fill_viridis_c() +
facet_wrap(~ species) +
ggtitle("Simulated data") +
theme(legend.position = "none")
p_simulated
Spatial Polya-gamma multinomial regression
Below is a DAG for the model
## GGDags
## https://cran.r-project.org/web/packages/ggdag/vignettes/intro-to-ggdag.html
# install.packages("dagitty")
# install.packages("ggdag")
library(dagitty)
library(ggdag)
library(cowplot)
library(tidyverse)
library(latex2exp)
## set coordinates for dag
coords <- tibble::tribble(
~name, ~x, ~y,
"Y", 3, 1,
"eta", 2, 1,
"locs", 2.5, 1.25,
"tau2", 1.5, 0.75,
"theta", 1.5, 1.25,
"beta", 1, 1,
"X", 0, 1,
"mu_beta", 0.5, 0.75,
"Sigma_beta", 0.5, 1.25
)
dag <- dagify(
Y ~ eta,
eta ~ beta + locs + theta + tau2,
beta ~ X + mu_beta + Sigma_beta,
# exposure = "X",
outcome = "Y",
coords = coords
)
dag_tidy <- dag %>%
tidy_dagitty(seed = 404) %>%
arrange(name) %>%
mutate(type = case_when(
name %in% c("X", "Y", "locs") ~ "data",
name %in% c("tau2", "theta") ~ "hyperparameter",
name %in% c("mu_beta", "Sigma_beta") ~ "prior",
TRUE ~ "parameter"
))
## manually rearrange the values
# dag %>%
# tidy_dagitty(seed = 404) %>%
# arrange(name)
dag_tidy %>%
ggplot(aes(x = x, y = y, xend = xend, yend = yend, color = type)) +
geom_dag_point() +
geom_dag_edges() +
geom_dag_text(
color = "black",
label = c(
TeX("$\\beta"),
TeX("$\\eta"),
"lat/lon",
TeX("$\\mu_{\\beta}$"),
TeX("$\\Sigma_{\\beta}$"),
TeX("$\\tau^2"),
TeX("$\\theta$"),
"X",
"Y"
)
) +
theme_dag() +
scale_color_viridis_d(begin = 0.9, end = 0.4) +
theme(legend.position = "bottom")
Let $\mathbf{y}i = (y{i, 1}, \ldots, y_{i, J})'$ be a $J$-dimensional vector of counts where $M_i = \sum_{j=1}^J y_{ij}$ is the total count and $\boldsymbol{\pi}i = ( \pi{i, 1}, \ldots, \pi_{i, J})'$ is a vector of probabilities with $\sum_{j=1}^J \pi_{i, j} = 1$. Then, the likelihood of $\mathbf{y}_i$ is given by
\begin{align}
[\mathbf{y}i | M_i, \boldsymbol{\pi}_i] & = \frac{M_i!} {\prod{j=1}^J y_{i, j}!} \pi_{i1}^{y_{i, 1}} \cdots \pi_{iJ}^{y_{i, J}}
(#eq:multinomial)
\end{align}
The canonical multinomial regression uses a soft-max link function where the $J$-dimensional probabilities are modeled in $\mathcal{R}^{J-1}$ with $J-1$ dimensional relative to a fixed reference category. Assigning latent variables $\boldsymbol{\eta}i = (\eta{i, 1}, \ldots, \eta_{i, J-1})'$ the softmax (multi-logit) function for $j = 1, \ldots, J-1$ is
\begin{align}
\pi_{ij} = \frac{e^{\eta_{ij}}} {1 + \sum_{j=1}^{J-1} e^{\eta_{ij}}}
\end{align}
where this can be interpreted in an $\mathcal{R}^{J}$ dimensional space with $\eta_{i,J} \equiv 0$. Multinomial regression assumes that given an $N \times q$-dimensional design matrix $\mathbf{X}$ for $j = 1, \ldots, J-1$, the latent parameter $\eta_{i, j} = \mathbf{X}i \boldsymbol{\beta}_j$. After assigning each $j = 1, \ldots, J-1$ a $\operatorname{N}(\boldsymbol{\mu}\beta, \boldsymbol{\Sigma}_\beta)$ prior, the posterior distribution is
\begin{align}
[\boldsymbol{\beta} | \mathbf{y}] & \propto \prod_{i=1}^N [\mathbf{y}i | \boldsymbol{\beta}] \prod{j=1}^{J-1} [\boldsymbol{\beta}_j].
\end{align}
The difficulty in evaluating the above posterior is that the distribution is not available in closed form and sampling requires a Metropolis-Hastings update (or some other non-conjugate sampler). This motivates the following data augmentation scheme.
The multinomial likelihood as a product of binomial distributions.
We can re-write the multinomial distribution in \@ref(eq:multinomial) as a recursive product of $J-1$ binomial distributions
\begin{align}
[\mathbf{y}i | M_i, \boldsymbol{\pi}_i] & = \operatorname{Mult} \left(M_i, \pi_i \right) \
& = \prod{j=1}^{J-1} \operatorname{Binomial} \left( y_{i,j} \middle| \widetilde{M}{i, j}, \widetilde{\pi}{i, j} \right) \
& = \prod_{j=1}^{J-1} \binom{\widetilde{M}{i, j}}{y{i, j}} \widetilde{\pi}{i, j}^{y{i, j}} (1 - \widetilde{\pi}{i, j})^{\widetilde{M}{i, j} - y_{i, j}}
\end{align}
where
\begin{align}
\widetilde{M}{i, j} & = \begin{cases}
\widetilde{M}{i, j} & \mbox{ if } j = 1 \
\widetilde{M}{i, j} - \sum{k < j} y_{i, k} & \mbox{ if } 1 < j \leq J - 1
\end{cases}
\end{align}
and the transformed (conditional) probabilities $\widetilde{\pi}_{i, j}$ recursively defined by
\begin{align}
\widetilde{\pi}{i, j} & = \begin{cases}
\pi{i, 1} & \mbox{ if } j = 1 \
\frac{\pi_{i, j}}{1 - \sum_{k < j} \pi_{i, k}} & \mbox{ if } 1 < j \leq J - 1
\end{cases}
\end{align}
where the stick-breaking transformation $\pi_{SB} \left( \boldsymbol{\eta}_{i} \right)$ maps the $J-1$ dimensional vector $\boldsymbol{\eta}_i$ over $\mathcal{R}^{J-1}$ to the $J$-dimensional unit simplex by
\begin{align}
\pi_{SB} \left( \eta_{i, j} \right) =
\frac{e^{ \eta_{i, j}} }{ \prod_{k \leq j} 1 + e^{ \eta_{i, j} } }.
\end{align}
Pólya-gamma data augmentation
The key idea for the Pólya-gamma data augmentation is that the multinomial likelihood can be written as
\begin{align}
[\mathbf{y}i | \boldsymbol{\eta}_i] & = \prod{j=1}^{J-1} \binom{\widetilde{M}{i, j}}{y{i, j}} \widetilde{\pi}{i, j}^{y{i, j}} (1 - \widetilde{\pi}{i, j})^{\widetilde{M}{i, j} - y_{i, j}} \nonumber \
& \propto \prod_{j=1}^{J-1}\frac{ (e^{\eta_{i,j}})^{a_{i, j}} }{(1 + e^{\eta_{i,j}})^{b_{i, j}}}
(#eq:likelihood)
\end{align}
where $\widetilde{\pi_{i,j}} = \frac{e^{\eta_{i,j}}}{1 + e^{\eta_{i,j}}}$ for some latent variable $\eta_{i, j}$ on the real line, $a_{i, j} = y_{i, j}$, and $b_{i, j} = \widetilde{M_{i, j}}$. Then, applying the identity [@polson2013bayesian]
\begin{align}
\frac{\left( e^{\eta_{i, j}} \right)^{y_{i, j}} }{ \left( 1 + e^{\eta_{i, j}} \right)^{\widetilde{M}{i, j} }} & = 2^{-\widetilde{M}{i, j}} e^{\kappa_{i, j} \eta_{i, j}} \int_0^\infty e^{- \omega_{i, j} \eta_{i, j}^2 / 2} \left[\omega_{i, j} | \widetilde{M}{i, j}, 0 \right] \,d\omega{i, j}
(#eq:pg-identity)
\end{align}
where $\kappa \left( y_{i, j} \right) = y_{i, j} - \widetilde{M}{i, j} / 2$.
Equation \@ref(eq:pg-identity) allows for the expression of the likelihood in \@ref(eq:likelihood) as an infinite convolution over the density $\left[\omega{i, j} | \widetilde{M}{i, j}, 0 \right]$ which is the probability density function of a Pólya-gamma random variable $\operatorname{PG} \left(\widetilde{M}{i, j}, 0 \right)$ and a component $e^{- \omega_{i, j} \eta_{i, j}^2 / 2}$ which is proportional to the kernel of a Gaussian density with precision $\omega_{i, j}$. We make the assumption that for all $i$ and $j$, $\omega_{i, j} \stackrel{iid}{\sim} \operatorname{PG} \left(\widetilde{M}{i, j}, 0 \right)$ Therefore, we can express a multinomial likelihood as an infinite convolution of a Gaussian random variable with a Pólya-gamma density. After defining a prior $[\boldsymbol{\eta}] = \prod{i=1}^N \prod_{j=1}^{J-1} [\eta_{i, j} | \boldsymbol{\eta}{-i, -j}]$ where $\boldsymbol{\eta}{-i, -j}$ is all of the elements of $\boldsymbol{\eta}$ except the $ij$th element, we can write the joint distribution $[\mathbf{y}, \boldsymbol{\eta}]$ as
Work on this notation
\begin{align}
[\mathbf{y}, \boldsymbol{\eta}] & \propto \prod_{i=1}^N \prod_{j=1}^{J-1}\frac{ (e^{\eta_{i,j}})^{y_{i, j}} }{(1 + e^{\eta_{i,j}})^{\widetilde{M}{i, j}}} [\boldsymbol{\eta}] \nonumber \
& \propto \prod{i=1}^N \prod_{j=1}^{J-1} 2^{-\widetilde{M}{i, j}} e^{\kappa(y{i, j}) \eta_{i, j}} \int_0^\infty e^{- \omega_{i, j} \eta_{i, j}^2 / 2} \left[\omega_{i, j} | \widetilde{M}{i, j}, 0 \right] \,d\omega{i, j} [\boldsymbol{\eta}] \nonumber \
& \propto \prod_{i=1}^N \prod_{j=1}^{J-1} \int_0^\infty [\eta_{i, j} | \boldsymbol{\eta}{-i,-j}] 2^{-\widetilde{M}{i, j}} e^{\kappa(y_{i, j}) \eta_{i, j}} e^{- \omega_{i, j} \eta_{i, j}^2 / 2} \left[\omega_{i, j} | \widetilde{M}{i, j}, 0 \right] \,d\omega{i, j} \nonumber \
& \propto \prod_{i=1}^N \prod_{j=1}^{J-1} \int_0^\infty [\mathbf{y}, \eta_{i, j}, \omega_{i, j} | \boldsymbol{\eta}{i, j}] \,d\omega{i, j} \nonumber \
& \propto \int_0^\infty [\mathbf{y}, \boldsymbol{\eta}, \boldsymbol{\omega}] \,d\boldsymbol{\omega}
(#eq:da)
\end{align}
where $[\mathbf{y}, \boldsymbol{\eta}, \boldsymbol{\omega}]$ is a joint density over the data augmented likelihood. When the prior on $\boldsymbol{\eta}$ is Gaussian, the marginal density $[\boldsymbol{\eta} | \mathbf{y}, \boldsymbol{\omega}] \propto \prod_{i=1}^N \prod_{j=1}^{J-1} e^{\kappa(y_{i, j}) \eta_{i, j}} e^{- \omega_{i, j} \eta_{i, j}^2 / 2} [\boldsymbol{\eta}]$ induced by the integrand in \@ref(eq:da) is also Gaussian. In addition, the exponential tilting property of the Pólya-gamma distribution [@polson2013bayesian] gives the conditional distribution
\begin{align}
[\omega_{i, j} | \mathbf{y}, \boldsymbol{\eta}] & \sim \operatorname{PG}(\widetilde{M}{i, j}, \eta{i, j})
\end{align}
Pólya-gamma regression
To perform regression on the multinomial vector given an $N \times q$ design matrix $\mathbf{X}$, we assume that $\eta_{i j} = \mathbf{X}i \boldsymbol{\beta}_j$ and $\boldsymbol{\beta}_j \sim \operatorname{N}(\boldsymbol{\mu}\beta, \boldsymbol{\Sigma}_\beta)$.
Defining $\boldsymbol{\Omega}i = \operatorname{diag}(\omega{i, 1}, \ldots, \omega_{i, J-1})$, we can calculate the full conditional distributions.
Full Conditionals
Full Conditional for $\boldsymbol{\beta}_{j}$
\begin{align}
\boldsymbol{\beta}{j} | \mathbf{y}, \boldsymbol{\omega} & \propto \prod{i=1}^N \operatorname{N} \left( \boldsymbol{\beta}{j} | \boldsymbol{\Omega}_i^{-1} \kappa \left( \mathbf{y}{i} \right), \boldsymbol{\Omega}j^{-1} \right) \operatorname{N} \left( \boldsymbol{\beta}{\cdot, j} | \boldsymbol{\mu}{\beta_j}, \boldsymbol{\Sigma}{\beta_j} \right) \
& \propto \operatorname{N} \left( \boldsymbol{\beta}_{\cdot, j} | \tilde{\boldsymbol{\mu}}_j, \tilde{\boldsymbol{\Sigma}}_j \right)
\end{align}
where
\begin{align}
\tilde{\boldsymbol{\mu}}j & = \tilde{\boldsymbol{\Sigma}}_j \left( {\boldsymbol{\Sigma}{\beta}}^{-1} \boldsymbol{\mu}\beta + \sum{i=1}^N \mathbf{x}i' \kappa \left( \mathbf{y}{i, j} \right) \right), \mbox{ and }\
\tilde{\boldsymbol{\Sigma}}j & = \left( {\boldsymbol{\Sigma}{\beta}}^{-1} + \sum_{i=1}^N \mathbf{x}i' \omega{i, j} \mathbf{x}_i \right)^{-1}
\end{align}
where $\mathbf{x}_i$ is the $i$th row of $\mathbf{X}$.
Full Conditional for $\omega_{i, j}$
If $\widetilde{M}{i, j} = 0$, then $\omega{i, j} | \mathbf{y}, \boldsymbol{\beta} \equiv 0$. Otherwise, for $\widetilde{M}_{i, j} > 0$, we have
\begin{align}
\omega_{i, j} | \mathbf{y}, \boldsymbol{\beta} & \propto \frac{e^{- \frac{1}{2} \omega_{i, j} \mathbf{x}i' \boldsymbol{\beta}_j}[\omega{i, j}]}{\int_{0}^{\infty} e^{- \frac{1}{2} \omega_{i, j} \mathbf{x}i' \boldsymbol{\beta}_j}[\omega{i, j}] \,d\omega_{i, j}}
%\operatorname{N} \left( \mathbf{y}{i} | \boldsymbol{\Omega}_i^{-1} \kappa \left( \mathbf{y}{i} \right), \boldsymbol{\Omega}i^{-1} \right) \operatorname{N} \left( \boldsymbol{\beta}{\cdot, j} | \boldsymbol{\mu}{\beta_j}, \boldsymbol{\Sigma}{\beta_j} \right) \
\end{align}
which is $\operatorname{PG} \left( \widetilde{M}{i, j}, \eta{i, j} \equiv \mathbf{x}_i' \boldsymbol{\beta}_j\right)$ by the exponential tilting property of the Pólya-gamma distribution.
params <- default_params()
params$n_adapt <- 500
params$n_mcmc <- 500
params$n_message <- 50
params$n_thin <- 1
priors <- default_priors_pg_splm(Y, X)
inits <- default_inits_pg_splm(Y, X, priors, shared_covariance_params = TRUE)
if (file.exists(here::here("results", "pg_splm-exponential-shared.RData"))) {
load(here::here("results", "pg_splm-exponential-shared.RData"))
} else {
## need to parallelize the polya-gamma random variables
## can I do this efficiently using openMP?
## Q: faster to parallelize using openMP or foreach?
start <- Sys.time()
out <- pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 32L, shared_covariance_params = TRUE)
stop <- Sys.time()
runtime <- stop - start
save(out, runtime, file = here::here("results", "pg_splm-exponential-shared.RData"))
}
layout(matrix(1:6, 3, 2))
for (i in 1:5) {
matplot(out$beta[, , i], type = 'l', main = paste("species", i))
abline(h = beta[, i], col = 1:nrow(beta))
}
matplot(out$theta, type = 'l', main = "theta")
abline(h = theta, col = 1:2)
layout(matrix(1:6, 3, 2))
plot_idx <- sample(1:N, 20)
for (i in 1:5) {
matplot(out$eta[, plot_idx, i], type = 'l', main = paste("species", i))
abline(h = eta[plot_idx, i], col = 1:20)
}
plot(out$tau2, type = 'l', main = "tau2")
abline(h = tau2, col = "red")
runtime
## plot beta estimates
dat_plot <- data.frame(
beta = c(
c(apply(out$beta, c(2, 3), mean)),
c(beta)
),
type = rep(c("estimate", "truth"), each = (J-1) * ncol(X)),
species = factor(rep(1:(J-1), each = ncol(X))),
variable = 1:ncol(X)
)
p1 <- dat_plot %>%
pivot_wider(names_from = type, values_from = beta) %>%
ggplot(aes(x = estimate, y = truth)) +
# scale_color_viridis_d(begin = 0, end = 0.8) +
geom_point(alpha = 0.5) +
facet_wrap(~ species, nrow = 8) +
geom_abline(intercept = 0, slope = 1, col = "red") +
ggtitle("Estimated vs. simulated beta")
## plot eeta estimates
dat_plot <- data.frame(
eta = c(
c(apply(out$eta, c(2, 3), mean)),
c(eta)
),
type = rep(c("estimate", "truth"), each = (J-1) * N),
species = factor(rep(1:(J-1), each = N)),
observations = 1:N
)
p2 <- dat_plot %>%
pivot_wider(names_from = type, values_from = eta) %>%
ggplot(aes(x = estimate, y = truth)) +
# scale_color_viridis_d(begin = 0, end = 0.8) +
geom_point(alpha = 0.5) +
facet_wrap(~ species, nrow = 8) +
geom_abline(intercept = 0, slope = 1, col = "red") +
ggtitle("Estimated vs. simulated eta")
p1 + p2
predictions
set.seed(111)
n <- 400
s <- sample(N, n)
Y_s <- Y[s, ]
Y_oos <- Y[-s, ]
X_s <- X[s, ]
X_oos <- X[-s, ]
eta_s <- eta[s, ]
eta_oos <- eta[-s, ]
pi_s <- pi[s, ]
pi_oos <- pi[-s, ]
locs_s <- locs[s, ]
locs_oos <- locs[-s, ]
Y_prop_s <- Y_prop[s, ]
Y_prop_oos <- Y_prop[-s, ]
## put data into data.frame for plotting
dat <- data.frame(
Y = c(Y_prop_s),
lon = c(locs_s[, 1]),
lat = c(locs_s[, 2]),
species = factor(rep(1:J, each = n)),
pi = c(pi_s)
)
p_observed <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) +
geom_raster() +
scale_fill_viridis_c() +
facet_wrap(~ species) +
ggtitle("Sampled simulated probabilities") +
theme(legend.position = "none")
p_observed
params <- default_params()
params$n_adapt <- 500
params$n_mcmc <- 500
params$n_message <- 50
params$n_thin <- 1
priors <- default_priors_pg_splm(Y_s, X_s)
inits <- default_inits_pg_splm(Y_s, X_s, priors, shared_covariance_params = TRUE)
if (file.exists(here::here("results", "pg_splm-exponential-sample-shared.RData"))) {
load(here::here("results", "pg_splm-exponential-sample-shared.RData"))
} else {
## need to parallelize the polya-gamma random variables
## can I do this efficiently using openMP?
## Q: faster to parallelize using openMP or foreach?
start <- Sys.time()
out <- pg_splm(Y_s, as.matrix(X_s), as.matrix(locs_s), params, priors, corr_fun = "exponential", n_cores = 32L, shared_covariance_params = TRUE)
stop <- Sys.time()
runtime <- stop - start
save(out, runtime, file = here::here("results", "pg_splm-exponential-sample-shared.RData"))
}
layout(matrix(1:6, 3, 2))
for (i in 1:5) {
matplot(out$beta[, , i], type = 'l', main = paste("species", i))
abline(h = beta[, i], col = 1:nrow(beta))
}
matplot(out$theta, type = 'l', main = "theta")
abline(h = theta, col = 1:2)
layout(matrix(1:6, 3, 2))
plot_idx <- sample(1:n, 20)
for (i in 1:5) {
matplot(out$eta[, plot_idx, i], type = 'l', main = paste("species", i))
abline(h = eta_s[plot_idx, i], col = 1:20)
}
plot(out$tau2, type = 'l', main = "tau2")
abline(h = tau2, col = "red")
runtime
## plot beta estimates
dat_plot <- data.frame(
beta = c(
c(apply(out$beta, c(2, 3), mean)),
c(beta)
),
type = rep(c("estimate", "truth"), each = (J-1) * ncol(X)),
species = factor(rep(1:(J-1), each = ncol(X))),
variable = 1:ncol(X)
)
p1 <- dat_plot %>%
pivot_wider(names_from = type, values_from = beta) %>%
ggplot(aes(x = estimate, y = truth)) +
# scale_color_viridis_d(begin = 0, end = 0.8) +
geom_point(alpha = 0.5) +
facet_wrap(~ species, nrow = 8) +
geom_abline(intercept = 0, slope = 1, col = "red") +
ggtitle("Estimated vs. simulated beta")
## plot eeta estimates
dat_plot <- data.frame(
eta = c(
c(apply(out$eta, c(2, 3), mean)),
c(eta_s)
),
type = rep(c("estimate", "truth"), each = (J-1) * n),
species = factor(rep(1:(J-1), each = n)),
observations = 1:n
)
p2 <- dat_plot %>%
pivot_wider(names_from = type, values_from = eta) %>%
ggplot(aes(x = estimate, y = truth)) +
# scale_color_viridis_d(begin = 0, end = 0.8) +
geom_point(alpha = 0.5) +
facet_wrap(~ species, nrow = 8) +
geom_abline(intercept = 0, slope = 1, col = "red") +
ggtitle("Estimated vs. simulated eta")
p1 + p2
if (file.exists(here::here("results", "pg_splm-exponential-sample-preds-shared.RData"))) {
load(here::here("results", "pg_splm-exponential-sample-preds-shared.RData"))
} else {
## currently this is somewhat slow but is embarassingly parallel
preds <- predict_pg_splm(out, X_s, X_oos, as.matrix(locs_s), as.matrix(locs_oos), corr_fun = "exponential", shared_covariance_params = TRUE)
save(preds, file = here::here("results", "pg_splm-exponential-sample-preds-shared.RData"))
}
pi_pred_mean <- apply(preds$pi, c(2, 3), mean)
eta_pred_mean <- apply(preds$eta, c(2, 3), mean)
layout(matrix(1:2, 1, 2))
plot(eta_oos, eta_pred_mean)
abline(0, 1, col = "red")
plot(pi_oos, pi_pred_mean)
abline(0, 1, col = "red")
Plot the simulated data
## put data into data.frame for plotting
dat <- data.frame(
Y = c(rbind(Y_prop_s, Y_prop_oos)),
lon = c(locs_s[, 1], locs_oos[, 1]),
lat = c(locs_s[, 2], locs_oos[, 2]),
species = factor(rep(1:J, each = N)),
pi = c(rbind(pi_s, pi_pred_mean)),
type = c(rep("observed", n), rep("predicted", N-n))
)
p_predicted <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) +
# p_predicted <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi, alpha = type)) +
geom_raster() +
scale_fill_viridis_c() +
facet_wrap(~ species) +
# scale_alpha_discrete(range = c(0.6, 1)) +
ggtitle("Predicted latent probability") +
theme(legend.position = "none")
p_simulated + p_observed + p_predicted
Each component gets its own covariance parameters
set.seed(11)
N <- 30^2
J <- 6
p <- 2
## setup the spatial process
locs <- expand.grid(
seq(0, 1, length = sqrt(N)),
seq(0, 1, length = sqrt(N))
)
D <- fields::rdist(locs)
## use the same parameters for each of the J-1 components
tau2 <- runif(J-1, 0.25, 1.5)
theta <- rnorm(J-1, log(0.1), 0.1)
Sigma <- array(0, dim = c(J-1, N, N))
psi <- matrix(0, N, J-1)
for (j in 1:(J-1)) {
Sigma[j, , ] <- tau2[j] * correlation_function(D, theta[j], corr_fun = "exponential")
psi[, j] <- c(mvnfast::rmvn(1, rep(0, N), Sigma[j, , ]))
}
## setup the fixed effects process
X <- cbind(1, matrix(runif(N*p), N, p))
beta <- matrix(rnorm((J-1) * ncol(X), 0, 0.25), ncol(X), (J-1))
## make the intercepts smaller to reduce stochastic ordering effect
beta[1, ] <- beta[1, ] - seq(from = 2, to = 0, length.out = J-1)
eta <- X %*% beta + psi
pi <- eta_to_pi(eta)
Y <- matrix(0, N, J)
for (i in 1:N) {
Y[i, ] <- rmultinom(1, 500, pi[i, ])
}
Y_prop <- counts_to_proportions(Y)
Plot the simulated data
## put data into data.frame for plotting
dat <- data.frame(
Y = c(Y_prop),
lon = locs[, 1],
lat = locs[, 2],
species = factor(rep(1:J, each = N)),
pi = c(pi)
)
p_simulated <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) +
geom_raster() +
scale_fill_viridis_c() +
facet_wrap(~ species) +
ggtitle("Simulated data")
p_simulated
fit the model
params <- default_params()
params$n_adapt <- 500
params$n_mcmc <- 500
params$n_message <- 50
params$n_thin <- 1
priors <- default_priors_pg_splm(Y, X)
inits <- default_inits_pg_splm(Y, X, priors, corr_fun = "exponential", shared_covariance_params = FALSE)
if (file.exists(here::here("results", "pg_splm-exponential.RData"))) {
load(here::here("results", "pg_splm-exponential.RData"))
} else {
## parallelize the Cholesky decomposition
## add in tryCatch checks for all Cholesky decompositions
## add in option to count the number of Cholesky decomposition failures and add a single warning message at the end of the MCMC and add in a stopping rule if the number of warnings (as a % of MCMC iterations) gets too large
## need to parallelize the polya-gamma random variables
## can I do this efficiently using openMP?
## Q: faster to parallelize using openMP or foreach?
start <- Sys.time()
out <- pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 32L, shared_covariance_params = FALSE, verbose = FALSE)
stop <- Sys.time()
runtime <- stop - start
save(out, runtime, file = here::here("results", "pg_splm-exponential.RData"))
}
layout(matrix(1:6, 3, 2))
for (i in 1:5) {
matplot(out$beta[, , i], type = 'l', main = paste("species", i), col = 1:nrow(beta))
abline(h = beta[, i], col = 1:nrow(beta))
}
layout(matrix(1:6, 3, 2))
for (i in 1:5) {
matplot(out$theta[, i], type = 'l', main = "theta")
abline(h = theta[i], col = 1:2)
}
matplot(out$tau2, type = 'l')
abline(h = tau2, col = 1:length(tau2))
layout(matrix(1:6, 3, 2))
plot_idx <- sample(1:N, 20)
for (i in 1:5) {
matplot(out$eta[, plot_idx, i], type = 'l', main = paste("species", i))
abline(h = eta[plot_idx, i], col = 1:20)
}
runtime
## plot beta estimates
dat_plot <- data.frame(
beta = c(
c(apply(out$beta, c(2, 3), mean)),
c(beta)
),
type = rep(c("estimate", "truth"), each = (J-1) * ncol(X)),
species = factor(rep(1:(J-1), each = ncol(X))),
variable = 1:ncol(X)
)
p1 <- dat_plot %>%
pivot_wider(names_from = type, values_from = beta) %>%
ggplot(aes(x = estimate, y = truth)) +
# scale_color_viridis_d(begin = 0, end = 0.8) +
geom_point(alpha = 0.5) +
facet_wrap(~ species, nrow = 8) +
geom_abline(intercept = 0, slope = 1, col = "red") +
ggtitle("Estimated vs. simulated beta")
## plot eeta estimates
dat_plot <- data.frame(
eta = c(
c(apply(out$eta, c(2, 3), mean)),
c(eta)
),
type = rep(c("estimate", "truth"), each = (J-1) * N),
species = factor(rep(1:(J-1), each = N)),
observations = 1:N
)
p2 <- dat_plot %>%
pivot_wider(names_from = type, values_from = eta) %>%
ggplot(aes(x = estimate, y = truth)) +
# scale_color_viridis_d(begin = 0, end = 0.8) +
geom_point(alpha = 0.5) +
facet_wrap(~ species, nrow = 8) +
geom_abline(intercept = 0, slope = 1, col = "red") +
ggtitle("Estimated vs. simulated eta")
p1 + p2
predictions
set.seed(111)
n <- 400
s <- sample(N, n)
Y_s <- Y[s, ]
Y_oos <- Y[-s, ]
X_s <- X[s, ]
X_oos <- X[-s, ]
eta_s <- eta[s, ]
eta_oos <- eta[-s, ]
pi_s <- pi[s, ]
pi_oos <- pi[-s, ]
locs_s <- locs[s, ]
locs_oos <- locs[-s, ]
Y_prop_s <- Y_prop[s, ]
Y_prop_oos <- Y_prop[-s, ]
## put data into data.frame for plotting
dat <- data.frame(
Y = c(Y_prop_s),
lon = c(locs_s[, 1]),
lat = c(locs_s[, 2]),
species = factor(rep(1:J, each = n)),
pi = c(pi_s)
)
p_observed <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) +
geom_raster() +
scale_fill_viridis_c() +
facet_wrap(~ species) +
ggtitle("Sampled simulated probabilities")
p_observed
params <- default_params()
params$n_adapt <- 500
params$n_mcmc <- 500
params$n_message <- 50
params$n_thin <- 1
priors <- default_priors_pg_splm(Y_s, X_s)
inits <- default_inits_pg_splm(Y_s, X_s, priors, corr_fun = "exponential", shared_covariance_params = FALSE)
if (file.exists(here::here("results", "pg_splm-exponential-sample.RData"))) {
load(here::here("results", "pg_splm-exponential-sample.RData"))
} else {
## need to parallelize the polya-gamma random variables
## can I do this efficiently using openMP?
## Q: faster to parallelize using openMP or foreach?
start <- Sys.time()
out <- pg_splm(Y_s, as.matrix(X_s), as.matrix(locs_s), params, priors, corr_fun = "exponential", n_cores = 32L, shared_covariance_params = FALSE)
stop <- Sys.time()
runtime <- stop - start
save(out, runtime, file = here::here("results", "pg_splm-exponential-sample.RData"))
}
layout(matrix(1:6, 3, 2))
for (i in 1:5) {
matplot(out$beta[, , i], type = 'l', main = paste("species", i), col = 1:nrow(beta))
abline(h = beta[, i], col = 1:nrow(beta))
}
layout(matrix(1:6, 3, 2))
for (i in 1:5) {
matplot(out$theta[, i], type = 'l', main = "theta")
abline(h = theta[i], col = 1:2)
}
matplot(out$tau2, type = 'l')
abline(h = tau2, col = 1:length(tau2))
layout(matrix(1:6, 3, 2))
plot_idx <- sample(1:n, 20)
for (i in 1:5) {
matplot(out$eta[, plot_idx, i], type = 'l', main = paste("species", i))
abline(h = eta[plot_idx, i], col = 1:20)
}
runtime
## plot beta estimates
dat_plot <- data.frame(
beta = c(
c(apply(out$beta, c(2, 3), mean)),
c(beta)
),
type = rep(c("estimate", "truth"), each = (J-1) * ncol(X)),
species = factor(rep(1:(J-1), each = ncol(X))),
variable = 1:ncol(X)
)
p1 <- dat_plot %>%
pivot_wider(names_from = type, values_from = beta) %>%
ggplot(aes(x = estimate, y = truth)) +
# scale_color_viridis_d(begin = 0, end = 0.8) +
geom_point(alpha = 0.5) +
facet_wrap(~ species, nrow = 8) +
geom_abline(intercept = 0, slope = 1, col = "red") +
ggtitle("Estimated vs. simulated beta")
## plot eeta estimates
dat_plot <- data.frame(
eta = c(
c(apply(out$eta, c(2, 3), mean)),
c(eta_s)
),
type = rep(c("estimate", "truth"), each = (J-1) * n),
species = factor(rep(1:(J-1), each = n)),
observations = 1:n
)
p2 <- dat_plot %>%
pivot_wider(names_from = type, values_from = eta) %>%
ggplot(aes(x = estimate, y = truth)) +
# scale_color_viridis_d(begin = 0, end = 0.8) +
geom_point(alpha = 0.5) +
facet_wrap(~ species, nrow = 8) +
geom_abline(intercept = 0, slope = 1, col = "red") +
ggtitle("Estimated vs. simulated eta")
p1 + p2
if (file.exists(here::here("results", "pg_splm-exponential-sample-preds.RData"))) {
load(here::here("results", "pg_splm-exponential-sample-preds.RData"))
} else {
## currently this is somewhat slow but is embarassingly parallel
preds <- predict_pg_splm(out, X_s, X_oos, as.matrix(locs_s), as.matrix(locs_oos), corr_fun = "exponential", shared_covariance_params = FALSE)
save(preds, file = here::here("results", "pg_splm-exponential-sample-preds.RData"))
}
pi_pred_mean <- apply(preds$pi, c(2, 3), mean)
eta_pred_mean <- apply(preds$eta, c(2, 3), mean)
layout(matrix(1:2, 1, 2))
plot(eta_oos, eta_pred_mean)
abline(0, 1, col = "red")
plot(pi_oos, pi_pred_mean)
abline(0, 1, col = "red")
Plot the simulated data
## put data into data.frame for plotting
dat <- data.frame(
Y = c(rbind(Y_prop_s, Y_prop_oos)),
lon = c(locs_s[, 1], locs_oos[, 1]),
lat = c(locs_s[, 2], locs_oos[, 2]),
species = factor(rep(1:J, each = N)),
pi = c(rbind(pi_s, pi_pred_mean)),
type = c(rep("observed", n), rep("predicted", N-n))
)
p_predicted <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) +
# p_predicted <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi, alpha = type)) +
geom_raster() +
scale_fill_viridis_c() +
facet_wrap(~ species) +
# scale_alpha_discrete(range = c(0.6, 1)) +
ggtitle("Predicted latent probability")
p_simulated + p_observed + p_predicted
Loo model selection example shared vs separate covariance parameters
load(here::here("results", "pg_splm-exponential-sample-shared.RData"))
ll_shared <- calc_ll_pg_splm(Y_s, X_s, out)
load(here::here("results", "pg_splm-exponential-sample.RData"))
ll_not_shared <- calc_ll_pg_splm(Y_s, X_s, out)
## loo library
library(loo)
loo_shared <- loo(t(ll_shared$ll), cores = 32L)
loo_not_shared <- loo(t(ll_not_shared$ll), cores = 32L)
comp <- loo_compare(loo_shared, loo_not_shared)
# prefers the not shared covariance parameter model
print(comp)
plot(loo_shared)
plot(loo_not_shared)
code profiling
## not run -- diagnostic profiling-exponential
## profile the shared covariance parameter models
params <- default_params()
params$n_adapt <- 50
params$n_mcmc <- 50
priors <- default_priors_pg_splm(Y, X)
inits <- default_inits_pg_splm(Y, X, priors, corr_fun = "exponential", shared_covariance_params = TRUE)
profvis::profvis(pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 1L, shared_covariance_params = TRUE))
profvis::profvis(pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 6L, shared_covariance_params = TRUE))
## profile the component-specific covariance parameter models
params <- default_params()
params$n_adapt <- 50
params$n_mcmc <- 50
priors <- default_priors_pg_splm(Y, X)
inits <- default_inits_pg_splm(Y, X, priors, corr_fun = "exponential" shared_covariance_params = FALSE)
profvis::profvis(pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 1L, shared_covariance_params = FALSE))
profvis::profvis(pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 6L, shared_covariance_params = FALSE))
jtipton25/pgR documentation built on July 8, 2022, 12:44 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
We are testing the spatial Polya-gamma linear model pg_splm()
knitr::opts_chunk$set(fig.width = 16, fig.height = 9) library(pgR) library(mvnfast) # library(MCMCpack) library(splines) library(tidyverse) library(patchwork)
simulate some data
set.seed(11) N <- 30^2 J <- 6 p <- 2 ## setup the spatial process locs <- expand.grid( seq(0, 1, length = sqrt(N)), seq(0, 1, length = sqrt(N)) ) D <- fields::rdist(locs) ## use the same parameters for each of the J-1 components tau2 <- 0.4 theta <- log(0.1) Sigma <- tau2 * correlation_function(D, theta, corr_fun = "exponential") psi <- t(rmvn(J-1, rep(0, N), Sigma)) ## setup the fixed effects process X <- cbind(1, matrix(runif(N*p), N, p)) beta <- matrix(rnorm((J-1) * ncol(X), 0, 0.25), ncol(X), (J-1)) ## make the intercepts smaller to reduce stochastic ordering effect beta[1, ] <- beta[1, ] - seq(from = 2, to = 0, length.out = J-1) eta <- X %*% beta + psi pi <- eta_to_pi(eta) Y <- matrix(0, N, J) for (i in 1:N) { Y[i, ] <- rmultinom(1, 500, pi[i, ]) } Y_prop <- counts_to_proportions(Y)
Plot the simulated data
## put data into data.frame for plotting dat <- data.frame( Y = c(Y_prop), lon = locs[, 1], lat = locs[, 2], species = factor(rep(1:J, each = N)), pi = c(pi) ) p_simulated <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) + geom_raster() + scale_fill_viridis_c() + facet_wrap(~ species) + ggtitle("Simulated data") + theme(legend.position = "none") p_simulated
Spatial Polya-gamma multinomial regression
Below is a DAG for the model
## GGDags ## https://cran.r-project.org/web/packages/ggdag/vignettes/intro-to-ggdag.html # install.packages("dagitty") # install.packages("ggdag") library(dagitty) library(ggdag) library(cowplot) library(tidyverse) library(latex2exp) ## set coordinates for dag coords <- tibble::tribble( ~name, ~x, ~y, "Y", 3, 1, "eta", 2, 1, "locs", 2.5, 1.25, "tau2", 1.5, 0.75, "theta", 1.5, 1.25, "beta", 1, 1, "X", 0, 1, "mu_beta", 0.5, 0.75, "Sigma_beta", 0.5, 1.25 ) dag <- dagify( Y ~ eta, eta ~ beta + locs + theta + tau2, beta ~ X + mu_beta + Sigma_beta, # exposure = "X", outcome = "Y", coords = coords ) dag_tidy <- dag %>% tidy_dagitty(seed = 404) %>% arrange(name) %>% mutate(type = case_when( name %in% c("X", "Y", "locs") ~ "data", name %in% c("tau2", "theta") ~ "hyperparameter", name %in% c("mu_beta", "Sigma_beta") ~ "prior", TRUE ~ "parameter" )) ## manually rearrange the values # dag %>% # tidy_dagitty(seed = 404) %>% # arrange(name) dag_tidy %>% ggplot(aes(x = x, y = y, xend = xend, yend = yend, color = type)) + geom_dag_point() + geom_dag_edges() + geom_dag_text( color = "black", label = c( TeX("$\\beta"), TeX("$\\eta"), "lat/lon", TeX("$\\mu_{\\beta}$"), TeX("$\\Sigma_{\\beta}$"), TeX("$\\tau^2"), TeX("$\\theta$"), "X", "Y" ) ) + theme_dag() + scale_color_viridis_d(begin = 0.9, end = 0.4) + theme(legend.position = "bottom")
Let $\mathbf{y}i = (y{i, 1}, \ldots, y_{i, J})'$ be a $J$-dimensional vector of counts where $M_i = \sum_{j=1}^J y_{ij}$ is the total count and $\boldsymbol{\pi}i = ( \pi{i, 1}, \ldots, \pi_{i, J})'$ is a vector of probabilities with $\sum_{j=1}^J \pi_{i, j} = 1$. Then, the likelihood of $\mathbf{y}_i$ is given by
\begin{align} [\mathbf{y}i | M_i, \boldsymbol{\pi}_i] & = \frac{M_i!} {\prod{j=1}^J y_{i, j}!} \pi_{i1}^{y_{i, 1}} \cdots \pi_{iJ}^{y_{i, J}} (#eq:multinomial) \end{align}
The canonical multinomial regression uses a soft-max link function where the $J$-dimensional probabilities are modeled in $\mathcal{R}^{J-1}$ with $J-1$ dimensional relative to a fixed reference category. Assigning latent variables $\boldsymbol{\eta}i = (\eta{i, 1}, \ldots, \eta_{i, J-1})'$ the softmax (multi-logit) function for $j = 1, \ldots, J-1$ is
\begin{align} \pi_{ij} = \frac{e^{\eta_{ij}}} {1 + \sum_{j=1}^{J-1} e^{\eta_{ij}}} \end{align}
where this can be interpreted in an $\mathcal{R}^{J}$ dimensional space with $\eta_{i,J} \equiv 0$. Multinomial regression assumes that given an $N \times q$-dimensional design matrix $\mathbf{X}$ for $j = 1, \ldots, J-1$, the latent parameter $\eta_{i, j} = \mathbf{X}i \boldsymbol{\beta}_j$. After assigning each $j = 1, \ldots, J-1$ a $\operatorname{N}(\boldsymbol{\mu}\beta, \boldsymbol{\Sigma}_\beta)$ prior, the posterior distribution is
\begin{align} [\boldsymbol{\beta} | \mathbf{y}] & \propto \prod_{i=1}^N [\mathbf{y}i | \boldsymbol{\beta}] \prod{j=1}^{J-1} [\boldsymbol{\beta}_j]. \end{align}
The difficulty in evaluating the above posterior is that the distribution is not available in closed form and sampling requires a Metropolis-Hastings update (or some other non-conjugate sampler). This motivates the following data augmentation scheme.
The multinomial likelihood as a product of binomial distributions.
We can re-write the multinomial distribution in \@ref(eq:multinomial) as a recursive product of $J-1$ binomial distributions
\begin{align} [\mathbf{y}i | M_i, \boldsymbol{\pi}_i] & = \operatorname{Mult} \left(M_i, \pi_i \right) \ & = \prod{j=1}^{J-1} \operatorname{Binomial} \left( y_{i,j} \middle| \widetilde{M}{i, j}, \widetilde{\pi}{i, j} \right) \ & = \prod_{j=1}^{J-1} \binom{\widetilde{M}{i, j}}{y{i, j}} \widetilde{\pi}{i, j}^{y{i, j}} (1 - \widetilde{\pi}{i, j})^{\widetilde{M}{i, j} - y_{i, j}} \end{align}
where
\begin{align} \widetilde{M}{i, j} & = \begin{cases} \widetilde{M}{i, j} & \mbox{ if } j = 1 \ \widetilde{M}{i, j} - \sum{k < j} y_{i, k} & \mbox{ if } 1 < j \leq J - 1 \end{cases} \end{align}
and the transformed (conditional) probabilities $\widetilde{\pi}_{i, j}$ recursively defined by
\begin{align} \widetilde{\pi}{i, j} & = \begin{cases} \pi{i, 1} & \mbox{ if } j = 1 \ \frac{\pi_{i, j}}{1 - \sum_{k < j} \pi_{i, k}} & \mbox{ if } 1 < j \leq J - 1 \end{cases} \end{align}
where the stick-breaking transformation $\pi_{SB} \left( \boldsymbol{\eta}_{i} \right)$ maps the $J-1$ dimensional vector $\boldsymbol{\eta}_i$ over $\mathcal{R}^{J-1}$ to the $J$-dimensional unit simplex by
\begin{align} \pi_{SB} \left( \eta_{i, j} \right) = \frac{e^{ \eta_{i, j}} }{ \prod_{k \leq j} 1 + e^{ \eta_{i, j} } }. \end{align}
Pólya-gamma data augmentation
The key idea for the Pólya-gamma data augmentation is that the multinomial likelihood can be written as
\begin{align} [\mathbf{y}i | \boldsymbol{\eta}_i] & = \prod{j=1}^{J-1} \binom{\widetilde{M}{i, j}}{y{i, j}} \widetilde{\pi}{i, j}^{y{i, j}} (1 - \widetilde{\pi}{i, j})^{\widetilde{M}{i, j} - y_{i, j}} \nonumber \ & \propto \prod_{j=1}^{J-1}\frac{ (e^{\eta_{i,j}})^{a_{i, j}} }{(1 + e^{\eta_{i,j}})^{b_{i, j}}} (#eq:likelihood) \end{align}
where $\widetilde{\pi_{i,j}} = \frac{e^{\eta_{i,j}}}{1 + e^{\eta_{i,j}}}$ for some latent variable $\eta_{i, j}$ on the real line, $a_{i, j} = y_{i, j}$, and $b_{i, j} = \widetilde{M_{i, j}}$. Then, applying the identity [@polson2013bayesian]
\begin{align} \frac{\left( e^{\eta_{i, j}} \right)^{y_{i, j}} }{ \left( 1 + e^{\eta_{i, j}} \right)^{\widetilde{M}{i, j} }} & = 2^{-\widetilde{M}{i, j}} e^{\kappa_{i, j} \eta_{i, j}} \int_0^\infty e^{- \omega_{i, j} \eta_{i, j}^2 / 2} \left[\omega_{i, j} | \widetilde{M}{i, j}, 0 \right] \,d\omega{i, j} (#eq:pg-identity) \end{align}
where $\kappa \left( y_{i, j} \right) = y_{i, j} - \widetilde{M}{i, j} / 2$. Equation \@ref(eq:pg-identity) allows for the expression of the likelihood in \@ref(eq:likelihood) as an infinite convolution over the density $\left[\omega{i, j} | \widetilde{M}{i, j}, 0 \right]$ which is the probability density function of a Pólya-gamma random variable $\operatorname{PG} \left(\widetilde{M}{i, j}, 0 \right)$ and a component $e^{- \omega_{i, j} \eta_{i, j}^2 / 2}$ which is proportional to the kernel of a Gaussian density with precision $\omega_{i, j}$. We make the assumption that for all $i$ and $j$, $\omega_{i, j} \stackrel{iid}{\sim} \operatorname{PG} \left(\widetilde{M}{i, j}, 0 \right)$ Therefore, we can express a multinomial likelihood as an infinite convolution of a Gaussian random variable with a Pólya-gamma density. After defining a prior $[\boldsymbol{\eta}] = \prod{i=1}^N \prod_{j=1}^{J-1} [\eta_{i, j} | \boldsymbol{\eta}{-i, -j}]$ where $\boldsymbol{\eta}{-i, -j}$ is all of the elements of $\boldsymbol{\eta}$ except the $ij$th element, we can write the joint distribution $[\mathbf{y}, \boldsymbol{\eta}]$ as
Work on this notation \begin{align} [\mathbf{y}, \boldsymbol{\eta}] & \propto \prod_{i=1}^N \prod_{j=1}^{J-1}\frac{ (e^{\eta_{i,j}})^{y_{i, j}} }{(1 + e^{\eta_{i,j}})^{\widetilde{M}{i, j}}} [\boldsymbol{\eta}] \nonumber \ & \propto \prod{i=1}^N \prod_{j=1}^{J-1} 2^{-\widetilde{M}{i, j}} e^{\kappa(y{i, j}) \eta_{i, j}} \int_0^\infty e^{- \omega_{i, j} \eta_{i, j}^2 / 2} \left[\omega_{i, j} | \widetilde{M}{i, j}, 0 \right] \,d\omega{i, j} [\boldsymbol{\eta}] \nonumber \ & \propto \prod_{i=1}^N \prod_{j=1}^{J-1} \int_0^\infty [\eta_{i, j} | \boldsymbol{\eta}{-i,-j}] 2^{-\widetilde{M}{i, j}} e^{\kappa(y_{i, j}) \eta_{i, j}} e^{- \omega_{i, j} \eta_{i, j}^2 / 2} \left[\omega_{i, j} | \widetilde{M}{i, j}, 0 \right] \,d\omega{i, j} \nonumber \ & \propto \prod_{i=1}^N \prod_{j=1}^{J-1} \int_0^\infty [\mathbf{y}, \eta_{i, j}, \omega_{i, j} | \boldsymbol{\eta}{i, j}] \,d\omega{i, j} \nonumber \ & \propto \int_0^\infty [\mathbf{y}, \boldsymbol{\eta}, \boldsymbol{\omega}] \,d\boldsymbol{\omega} (#eq:da) \end{align}
where $[\mathbf{y}, \boldsymbol{\eta}, \boldsymbol{\omega}]$ is a joint density over the data augmented likelihood. When the prior on $\boldsymbol{\eta}$ is Gaussian, the marginal density $[\boldsymbol{\eta} | \mathbf{y}, \boldsymbol{\omega}] \propto \prod_{i=1}^N \prod_{j=1}^{J-1} e^{\kappa(y_{i, j}) \eta_{i, j}} e^{- \omega_{i, j} \eta_{i, j}^2 / 2} [\boldsymbol{\eta}]$ induced by the integrand in \@ref(eq:da) is also Gaussian. In addition, the exponential tilting property of the Pólya-gamma distribution [@polson2013bayesian] gives the conditional distribution
\begin{align} [\omega_{i, j} | \mathbf{y}, \boldsymbol{\eta}] & \sim \operatorname{PG}(\widetilde{M}{i, j}, \eta{i, j}) \end{align}
Pólya-gamma regression
To perform regression on the multinomial vector given an $N \times q$ design matrix $\mathbf{X}$, we assume that $\eta_{i j} = \mathbf{X}i \boldsymbol{\beta}_j$ and $\boldsymbol{\beta}_j \sim \operatorname{N}(\boldsymbol{\mu}\beta, \boldsymbol{\Sigma}_\beta)$.
Defining $\boldsymbol{\Omega}i = \operatorname{diag}(\omega{i, 1}, \ldots, \omega_{i, J-1})$, we can calculate the full conditional distributions.
Full Conditionals
Full Conditional for $\boldsymbol{\beta}_{j}$
\begin{align} \boldsymbol{\beta}{j} | \mathbf{y}, \boldsymbol{\omega} & \propto \prod{i=1}^N \operatorname{N} \left( \boldsymbol{\beta}{j} | \boldsymbol{\Omega}_i^{-1} \kappa \left( \mathbf{y}{i} \right), \boldsymbol{\Omega}j^{-1} \right) \operatorname{N} \left( \boldsymbol{\beta}{\cdot, j} | \boldsymbol{\mu}{\beta_j}, \boldsymbol{\Sigma}{\beta_j} \right) \ & \propto \operatorname{N} \left( \boldsymbol{\beta}_{\cdot, j} | \tilde{\boldsymbol{\mu}}_j, \tilde{\boldsymbol{\Sigma}}_j \right) \end{align}
where
\begin{align} \tilde{\boldsymbol{\mu}}j & = \tilde{\boldsymbol{\Sigma}}_j \left( {\boldsymbol{\Sigma}{\beta}}^{-1} \boldsymbol{\mu}\beta + \sum{i=1}^N \mathbf{x}i' \kappa \left( \mathbf{y}{i, j} \right) \right), \mbox{ and }\ \tilde{\boldsymbol{\Sigma}}j & = \left( {\boldsymbol{\Sigma}{\beta}}^{-1} + \sum_{i=1}^N \mathbf{x}i' \omega{i, j} \mathbf{x}_i \right)^{-1} \end{align}
where $\mathbf{x}_i$ is the $i$th row of $\mathbf{X}$.
Full Conditional for $\omega_{i, j}$
If $\widetilde{M}{i, j} = 0$, then $\omega{i, j} | \mathbf{y}, \boldsymbol{\beta} \equiv 0$. Otherwise, for $\widetilde{M}_{i, j} > 0$, we have
\begin{align} \omega_{i, j} | \mathbf{y}, \boldsymbol{\beta} & \propto \frac{e^{- \frac{1}{2} \omega_{i, j} \mathbf{x}i' \boldsymbol{\beta}_j}[\omega{i, j}]}{\int_{0}^{\infty} e^{- \frac{1}{2} \omega_{i, j} \mathbf{x}i' \boldsymbol{\beta}_j}[\omega{i, j}] \,d\omega_{i, j}} %\operatorname{N} \left( \mathbf{y}{i} | \boldsymbol{\Omega}_i^{-1} \kappa \left( \mathbf{y}{i} \right), \boldsymbol{\Omega}i^{-1} \right) \operatorname{N} \left( \boldsymbol{\beta}{\cdot, j} | \boldsymbol{\mu}{\beta_j}, \boldsymbol{\Sigma}{\beta_j} \right) \ \end{align}
which is $\operatorname{PG} \left( \widetilde{M}{i, j}, \eta{i, j} \equiv \mathbf{x}_i' \boldsymbol{\beta}_j\right)$ by the exponential tilting property of the Pólya-gamma distribution.
params <- default_params() params$n_adapt <- 500 params$n_mcmc <- 500 params$n_message <- 50 params$n_thin <- 1 priors <- default_priors_pg_splm(Y, X) inits <- default_inits_pg_splm(Y, X, priors, shared_covariance_params = TRUE) if (file.exists(here::here("results", "pg_splm-exponential-shared.RData"))) { load(here::here("results", "pg_splm-exponential-shared.RData")) } else { ## need to parallelize the polya-gamma random variables ## can I do this efficiently using openMP? ## Q: faster to parallelize using openMP or foreach? start <- Sys.time() out <- pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 32L, shared_covariance_params = TRUE) stop <- Sys.time() runtime <- stop - start save(out, runtime, file = here::here("results", "pg_splm-exponential-shared.RData")) }
layout(matrix(1:6, 3, 2)) for (i in 1:5) { matplot(out$beta[, , i], type = 'l', main = paste("species", i)) abline(h = beta[, i], col = 1:nrow(beta)) } matplot(out$theta, type = 'l', main = "theta") abline(h = theta, col = 1:2)
layout(matrix(1:6, 3, 2)) plot_idx <- sample(1:N, 20) for (i in 1:5) { matplot(out$eta[, plot_idx, i], type = 'l', main = paste("species", i)) abline(h = eta[plot_idx, i], col = 1:20) } plot(out$tau2, type = 'l', main = "tau2") abline(h = tau2, col = "red")
runtime ## plot beta estimates dat_plot <- data.frame( beta = c( c(apply(out$beta, c(2, 3), mean)), c(beta) ), type = rep(c("estimate", "truth"), each = (J-1) * ncol(X)), species = factor(rep(1:(J-1), each = ncol(X))), variable = 1:ncol(X) ) p1 <- dat_plot %>% pivot_wider(names_from = type, values_from = beta) %>% ggplot(aes(x = estimate, y = truth)) + # scale_color_viridis_d(begin = 0, end = 0.8) + geom_point(alpha = 0.5) + facet_wrap(~ species, nrow = 8) + geom_abline(intercept = 0, slope = 1, col = "red") + ggtitle("Estimated vs. simulated beta") ## plot eeta estimates dat_plot <- data.frame( eta = c( c(apply(out$eta, c(2, 3), mean)), c(eta) ), type = rep(c("estimate", "truth"), each = (J-1) * N), species = factor(rep(1:(J-1), each = N)), observations = 1:N ) p2 <- dat_plot %>% pivot_wider(names_from = type, values_from = eta) %>% ggplot(aes(x = estimate, y = truth)) + # scale_color_viridis_d(begin = 0, end = 0.8) + geom_point(alpha = 0.5) + facet_wrap(~ species, nrow = 8) + geom_abline(intercept = 0, slope = 1, col = "red") + ggtitle("Estimated vs. simulated eta") p1 + p2
predictions
set.seed(111) n <- 400 s <- sample(N, n) Y_s <- Y[s, ] Y_oos <- Y[-s, ] X_s <- X[s, ] X_oos <- X[-s, ] eta_s <- eta[s, ] eta_oos <- eta[-s, ] pi_s <- pi[s, ] pi_oos <- pi[-s, ] locs_s <- locs[s, ] locs_oos <- locs[-s, ] Y_prop_s <- Y_prop[s, ] Y_prop_oos <- Y_prop[-s, ] ## put data into data.frame for plotting dat <- data.frame( Y = c(Y_prop_s), lon = c(locs_s[, 1]), lat = c(locs_s[, 2]), species = factor(rep(1:J, each = n)), pi = c(pi_s) ) p_observed <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) + geom_raster() + scale_fill_viridis_c() + facet_wrap(~ species) + ggtitle("Sampled simulated probabilities") + theme(legend.position = "none") p_observed
params <- default_params() params$n_adapt <- 500 params$n_mcmc <- 500 params$n_message <- 50 params$n_thin <- 1 priors <- default_priors_pg_splm(Y_s, X_s) inits <- default_inits_pg_splm(Y_s, X_s, priors, shared_covariance_params = TRUE) if (file.exists(here::here("results", "pg_splm-exponential-sample-shared.RData"))) { load(here::here("results", "pg_splm-exponential-sample-shared.RData")) } else { ## need to parallelize the polya-gamma random variables ## can I do this efficiently using openMP? ## Q: faster to parallelize using openMP or foreach? start <- Sys.time() out <- pg_splm(Y_s, as.matrix(X_s), as.matrix(locs_s), params, priors, corr_fun = "exponential", n_cores = 32L, shared_covariance_params = TRUE) stop <- Sys.time() runtime <- stop - start save(out, runtime, file = here::here("results", "pg_splm-exponential-sample-shared.RData")) }
layout(matrix(1:6, 3, 2)) for (i in 1:5) { matplot(out$beta[, , i], type = 'l', main = paste("species", i)) abline(h = beta[, i], col = 1:nrow(beta)) } matplot(out$theta, type = 'l', main = "theta") abline(h = theta, col = 1:2)
layout(matrix(1:6, 3, 2)) plot_idx <- sample(1:n, 20) for (i in 1:5) { matplot(out$eta[, plot_idx, i], type = 'l', main = paste("species", i)) abline(h = eta_s[plot_idx, i], col = 1:20) } plot(out$tau2, type = 'l', main = "tau2") abline(h = tau2, col = "red")
runtime ## plot beta estimates dat_plot <- data.frame( beta = c( c(apply(out$beta, c(2, 3), mean)), c(beta) ), type = rep(c("estimate", "truth"), each = (J-1) * ncol(X)), species = factor(rep(1:(J-1), each = ncol(X))), variable = 1:ncol(X) ) p1 <- dat_plot %>% pivot_wider(names_from = type, values_from = beta) %>% ggplot(aes(x = estimate, y = truth)) + # scale_color_viridis_d(begin = 0, end = 0.8) + geom_point(alpha = 0.5) + facet_wrap(~ species, nrow = 8) + geom_abline(intercept = 0, slope = 1, col = "red") + ggtitle("Estimated vs. simulated beta") ## plot eeta estimates dat_plot <- data.frame( eta = c( c(apply(out$eta, c(2, 3), mean)), c(eta_s) ), type = rep(c("estimate", "truth"), each = (J-1) * n), species = factor(rep(1:(J-1), each = n)), observations = 1:n ) p2 <- dat_plot %>% pivot_wider(names_from = type, values_from = eta) %>% ggplot(aes(x = estimate, y = truth)) + # scale_color_viridis_d(begin = 0, end = 0.8) + geom_point(alpha = 0.5) + facet_wrap(~ species, nrow = 8) + geom_abline(intercept = 0, slope = 1, col = "red") + ggtitle("Estimated vs. simulated eta") p1 + p2
if (file.exists(here::here("results", "pg_splm-exponential-sample-preds-shared.RData"))) { load(here::here("results", "pg_splm-exponential-sample-preds-shared.RData")) } else { ## currently this is somewhat slow but is embarassingly parallel preds <- predict_pg_splm(out, X_s, X_oos, as.matrix(locs_s), as.matrix(locs_oos), corr_fun = "exponential", shared_covariance_params = TRUE) save(preds, file = here::here("results", "pg_splm-exponential-sample-preds-shared.RData")) }
pi_pred_mean <- apply(preds$pi, c(2, 3), mean) eta_pred_mean <- apply(preds$eta, c(2, 3), mean) layout(matrix(1:2, 1, 2)) plot(eta_oos, eta_pred_mean) abline(0, 1, col = "red") plot(pi_oos, pi_pred_mean) abline(0, 1, col = "red")
Plot the simulated data
## put data into data.frame for plotting dat <- data.frame( Y = c(rbind(Y_prop_s, Y_prop_oos)), lon = c(locs_s[, 1], locs_oos[, 1]), lat = c(locs_s[, 2], locs_oos[, 2]), species = factor(rep(1:J, each = N)), pi = c(rbind(pi_s, pi_pred_mean)), type = c(rep("observed", n), rep("predicted", N-n)) ) p_predicted <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) + # p_predicted <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi, alpha = type)) + geom_raster() + scale_fill_viridis_c() + facet_wrap(~ species) + # scale_alpha_discrete(range = c(0.6, 1)) + ggtitle("Predicted latent probability") + theme(legend.position = "none") p_simulated + p_observed + p_predicted
Each component gets its own covariance parameters
set.seed(11) N <- 30^2 J <- 6 p <- 2 ## setup the spatial process locs <- expand.grid( seq(0, 1, length = sqrt(N)), seq(0, 1, length = sqrt(N)) ) D <- fields::rdist(locs) ## use the same parameters for each of the J-1 components tau2 <- runif(J-1, 0.25, 1.5) theta <- rnorm(J-1, log(0.1), 0.1) Sigma <- array(0, dim = c(J-1, N, N)) psi <- matrix(0, N, J-1) for (j in 1:(J-1)) { Sigma[j, , ] <- tau2[j] * correlation_function(D, theta[j], corr_fun = "exponential") psi[, j] <- c(mvnfast::rmvn(1, rep(0, N), Sigma[j, , ])) } ## setup the fixed effects process X <- cbind(1, matrix(runif(N*p), N, p)) beta <- matrix(rnorm((J-1) * ncol(X), 0, 0.25), ncol(X), (J-1)) ## make the intercepts smaller to reduce stochastic ordering effect beta[1, ] <- beta[1, ] - seq(from = 2, to = 0, length.out = J-1) eta <- X %*% beta + psi pi <- eta_to_pi(eta) Y <- matrix(0, N, J) for (i in 1:N) { Y[i, ] <- rmultinom(1, 500, pi[i, ]) } Y_prop <- counts_to_proportions(Y)
Plot the simulated data
## put data into data.frame for plotting dat <- data.frame( Y = c(Y_prop), lon = locs[, 1], lat = locs[, 2], species = factor(rep(1:J, each = N)), pi = c(pi) ) p_simulated <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) + geom_raster() + scale_fill_viridis_c() + facet_wrap(~ species) + ggtitle("Simulated data") p_simulated
fit the model
params <- default_params() params$n_adapt <- 500 params$n_mcmc <- 500 params$n_message <- 50 params$n_thin <- 1 priors <- default_priors_pg_splm(Y, X) inits <- default_inits_pg_splm(Y, X, priors, corr_fun = "exponential", shared_covariance_params = FALSE) if (file.exists(here::here("results", "pg_splm-exponential.RData"))) { load(here::here("results", "pg_splm-exponential.RData")) } else { ## parallelize the Cholesky decomposition ## add in tryCatch checks for all Cholesky decompositions ## add in option to count the number of Cholesky decomposition failures and add a single warning message at the end of the MCMC and add in a stopping rule if the number of warnings (as a % of MCMC iterations) gets too large ## need to parallelize the polya-gamma random variables ## can I do this efficiently using openMP? ## Q: faster to parallelize using openMP or foreach? start <- Sys.time() out <- pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 32L, shared_covariance_params = FALSE, verbose = FALSE) stop <- Sys.time() runtime <- stop - start save(out, runtime, file = here::here("results", "pg_splm-exponential.RData")) }
layout(matrix(1:6, 3, 2)) for (i in 1:5) { matplot(out$beta[, , i], type = 'l', main = paste("species", i), col = 1:nrow(beta)) abline(h = beta[, i], col = 1:nrow(beta)) }
layout(matrix(1:6, 3, 2)) for (i in 1:5) { matplot(out$theta[, i], type = 'l', main = "theta") abline(h = theta[i], col = 1:2) } matplot(out$tau2, type = 'l') abline(h = tau2, col = 1:length(tau2))
layout(matrix(1:6, 3, 2)) plot_idx <- sample(1:N, 20) for (i in 1:5) { matplot(out$eta[, plot_idx, i], type = 'l', main = paste("species", i)) abline(h = eta[plot_idx, i], col = 1:20) }
runtime ## plot beta estimates dat_plot <- data.frame( beta = c( c(apply(out$beta, c(2, 3), mean)), c(beta) ), type = rep(c("estimate", "truth"), each = (J-1) * ncol(X)), species = factor(rep(1:(J-1), each = ncol(X))), variable = 1:ncol(X) ) p1 <- dat_plot %>% pivot_wider(names_from = type, values_from = beta) %>% ggplot(aes(x = estimate, y = truth)) + # scale_color_viridis_d(begin = 0, end = 0.8) + geom_point(alpha = 0.5) + facet_wrap(~ species, nrow = 8) + geom_abline(intercept = 0, slope = 1, col = "red") + ggtitle("Estimated vs. simulated beta") ## plot eeta estimates dat_plot <- data.frame( eta = c( c(apply(out$eta, c(2, 3), mean)), c(eta) ), type = rep(c("estimate", "truth"), each = (J-1) * N), species = factor(rep(1:(J-1), each = N)), observations = 1:N ) p2 <- dat_plot %>% pivot_wider(names_from = type, values_from = eta) %>% ggplot(aes(x = estimate, y = truth)) + # scale_color_viridis_d(begin = 0, end = 0.8) + geom_point(alpha = 0.5) + facet_wrap(~ species, nrow = 8) + geom_abline(intercept = 0, slope = 1, col = "red") + ggtitle("Estimated vs. simulated eta") p1 + p2
predictions
set.seed(111) n <- 400 s <- sample(N, n) Y_s <- Y[s, ] Y_oos <- Y[-s, ] X_s <- X[s, ] X_oos <- X[-s, ] eta_s <- eta[s, ] eta_oos <- eta[-s, ] pi_s <- pi[s, ] pi_oos <- pi[-s, ] locs_s <- locs[s, ] locs_oos <- locs[-s, ] Y_prop_s <- Y_prop[s, ] Y_prop_oos <- Y_prop[-s, ] ## put data into data.frame for plotting dat <- data.frame( Y = c(Y_prop_s), lon = c(locs_s[, 1]), lat = c(locs_s[, 2]), species = factor(rep(1:J, each = n)), pi = c(pi_s) ) p_observed <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) + geom_raster() + scale_fill_viridis_c() + facet_wrap(~ species) + ggtitle("Sampled simulated probabilities") p_observed
params <- default_params() params$n_adapt <- 500 params$n_mcmc <- 500 params$n_message <- 50 params$n_thin <- 1 priors <- default_priors_pg_splm(Y_s, X_s) inits <- default_inits_pg_splm(Y_s, X_s, priors, corr_fun = "exponential", shared_covariance_params = FALSE) if (file.exists(here::here("results", "pg_splm-exponential-sample.RData"))) { load(here::here("results", "pg_splm-exponential-sample.RData")) } else { ## need to parallelize the polya-gamma random variables ## can I do this efficiently using openMP? ## Q: faster to parallelize using openMP or foreach? start <- Sys.time() out <- pg_splm(Y_s, as.matrix(X_s), as.matrix(locs_s), params, priors, corr_fun = "exponential", n_cores = 32L, shared_covariance_params = FALSE) stop <- Sys.time() runtime <- stop - start save(out, runtime, file = here::here("results", "pg_splm-exponential-sample.RData")) }
layout(matrix(1:6, 3, 2)) for (i in 1:5) { matplot(out$beta[, , i], type = 'l', main = paste("species", i), col = 1:nrow(beta)) abline(h = beta[, i], col = 1:nrow(beta)) }
layout(matrix(1:6, 3, 2)) for (i in 1:5) { matplot(out$theta[, i], type = 'l', main = "theta") abline(h = theta[i], col = 1:2) } matplot(out$tau2, type = 'l') abline(h = tau2, col = 1:length(tau2))
layout(matrix(1:6, 3, 2)) plot_idx <- sample(1:n, 20) for (i in 1:5) { matplot(out$eta[, plot_idx, i], type = 'l', main = paste("species", i)) abline(h = eta[plot_idx, i], col = 1:20) }
runtime ## plot beta estimates dat_plot <- data.frame( beta = c( c(apply(out$beta, c(2, 3), mean)), c(beta) ), type = rep(c("estimate", "truth"), each = (J-1) * ncol(X)), species = factor(rep(1:(J-1), each = ncol(X))), variable = 1:ncol(X) ) p1 <- dat_plot %>% pivot_wider(names_from = type, values_from = beta) %>% ggplot(aes(x = estimate, y = truth)) + # scale_color_viridis_d(begin = 0, end = 0.8) + geom_point(alpha = 0.5) + facet_wrap(~ species, nrow = 8) + geom_abline(intercept = 0, slope = 1, col = "red") + ggtitle("Estimated vs. simulated beta") ## plot eeta estimates dat_plot <- data.frame( eta = c( c(apply(out$eta, c(2, 3), mean)), c(eta_s) ), type = rep(c("estimate", "truth"), each = (J-1) * n), species = factor(rep(1:(J-1), each = n)), observations = 1:n ) p2 <- dat_plot %>% pivot_wider(names_from = type, values_from = eta) %>% ggplot(aes(x = estimate, y = truth)) + # scale_color_viridis_d(begin = 0, end = 0.8) + geom_point(alpha = 0.5) + facet_wrap(~ species, nrow = 8) + geom_abline(intercept = 0, slope = 1, col = "red") + ggtitle("Estimated vs. simulated eta") p1 + p2
if (file.exists(here::here("results", "pg_splm-exponential-sample-preds.RData"))) { load(here::here("results", "pg_splm-exponential-sample-preds.RData")) } else { ## currently this is somewhat slow but is embarassingly parallel preds <- predict_pg_splm(out, X_s, X_oos, as.matrix(locs_s), as.matrix(locs_oos), corr_fun = "exponential", shared_covariance_params = FALSE) save(preds, file = here::here("results", "pg_splm-exponential-sample-preds.RData")) }
pi_pred_mean <- apply(preds$pi, c(2, 3), mean) eta_pred_mean <- apply(preds$eta, c(2, 3), mean) layout(matrix(1:2, 1, 2)) plot(eta_oos, eta_pred_mean) abline(0, 1, col = "red") plot(pi_oos, pi_pred_mean) abline(0, 1, col = "red")
Plot the simulated data
## put data into data.frame for plotting dat <- data.frame( Y = c(rbind(Y_prop_s, Y_prop_oos)), lon = c(locs_s[, 1], locs_oos[, 1]), lat = c(locs_s[, 2], locs_oos[, 2]), species = factor(rep(1:J, each = N)), pi = c(rbind(pi_s, pi_pred_mean)), type = c(rep("observed", n), rep("predicted", N-n)) ) p_predicted <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi)) + # p_predicted <- ggplot(data = dat, aes(x = lon, y = lat, fill = pi, alpha = type)) + geom_raster() + scale_fill_viridis_c() + facet_wrap(~ species) + # scale_alpha_discrete(range = c(0.6, 1)) + ggtitle("Predicted latent probability") p_simulated + p_observed + p_predicted
Loo model selection example shared vs separate covariance parameters
load(here::here("results", "pg_splm-exponential-sample-shared.RData")) ll_shared <- calc_ll_pg_splm(Y_s, X_s, out) load(here::here("results", "pg_splm-exponential-sample.RData")) ll_not_shared <- calc_ll_pg_splm(Y_s, X_s, out) ## loo library library(loo) loo_shared <- loo(t(ll_shared$ll), cores = 32L) loo_not_shared <- loo(t(ll_not_shared$ll), cores = 32L) comp <- loo_compare(loo_shared, loo_not_shared) # prefers the not shared covariance parameter model print(comp) plot(loo_shared) plot(loo_not_shared)
code profiling
## not run -- diagnostic profiling-exponential ## profile the shared covariance parameter models params <- default_params() params$n_adapt <- 50 params$n_mcmc <- 50 priors <- default_priors_pg_splm(Y, X) inits <- default_inits_pg_splm(Y, X, priors, corr_fun = "exponential", shared_covariance_params = TRUE) profvis::profvis(pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 1L, shared_covariance_params = TRUE)) profvis::profvis(pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 6L, shared_covariance_params = TRUE)) ## profile the component-specific covariance parameter models params <- default_params() params$n_adapt <- 50 params$n_mcmc <- 50 priors <- default_priors_pg_splm(Y, X) inits <- default_inits_pg_splm(Y, X, priors, corr_fun = "exponential" shared_covariance_params = FALSE) profvis::profvis(pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 1L, shared_covariance_params = FALSE)) profvis::profvis(pg_splm(Y, as.matrix(X), as.matrix(locs), params, priors, corr_fun = "exponential", n_cores = 6L, shared_covariance_params = FALSE))
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.