From 9780c5c0e65c67ecc95207060fcbe5b08e0f695e Mon Sep 17 00:00:00 2001
From: Riku-Laine <28960190+Riku-Laine@users.noreply.github.com>
Date: Mon, 20 Jan 2020 11:54:04 +0200
Subject: [PATCH] Updated stan code
---
.../code_linear_dependency.stan | 53 +++++++++++--------
1 file changed, 30 insertions(+), 23 deletions(-)
diff --git a/analysis_and_scripts/code_linear_dependency.stan b/analysis_and_scripts/code_linear_dependency.stan
index a85bded..71298d3 100644
--- a/analysis_and_scripts/code_linear_dependency.stan
+++ b/analysis_and_scripts/code_linear_dependency.stan
@@ -1,7 +1,7 @@
data {
- int<lower=1> D; // Dimensions of the features an coefficient vectors
- int<lower=0> N_obs; // Number of "observed observations" (with T = 1)
- int<lower=0> N_cens; // Number of "censored observations" (with T = 0)
+ int<lower=1> D; // Dimensions of the features and coefficient vectors
+ int<lower=0> N_obs; // Number of observed observations (with T = 1)
+ int<lower=0> N_cens; // Number of censored observations (with T = 0)
int<lower=0> M; // Number of judges
real<lower=0> sigma_tau; // Prior for the variance parameters.
int<lower=1, upper=M> jj_obs[N_obs]; // judge_ID array
@@ -18,11 +18,12 @@ parameters {
vector[N_obs] Z_obs;
vector[N_cens] Z_cens;
- // Intercepts
- vector<lower=0>[M] sigma_T;
+ // Intercepts and their variance parameters.
+ real<lower=0> sigma_T;
real<lower=0> sigma_Y;
vector[M] alpha_T_raw;
+ // vector[M] alpha_T;
real alpha_Y_raw;
// Temporary variables to compute the coefficients
@@ -31,10 +32,8 @@ parameters {
real<lower=0> a_ZT; // Presume latent variable has a positive coefficient.
real<lower=0> a_ZY;
- real<lower=0> tau_XT;
- real<lower=0> tau_XY;
- real<lower=0> tau_ZT;
- real<lower=0> tau_ZY;
+ real<lower=0> tau_X; // RL 08DEC2019: Combine taus for X.
+ real<lower=0> tau_Z; // RL 24OCT2019: Combine taus for Z.
}
@@ -49,13 +48,18 @@ transformed parameters {
// Intercepts
vector[M] alpha_T;
real alpha_Y;
-
- beta_XT = a_XT / sqrt(tau_XT);
- beta_XY = a_XY / sqrt(tau_XY);
- beta_ZT = a_ZT / sqrt(tau_ZT);
- beta_ZY = a_ZY / sqrt(tau_ZY);
+
+ beta_XT = a_XT / sqrt(tau_X); // RL 08DEC2019: Combine taus for X.
+ beta_XY = a_XY / sqrt(tau_X);
+ beta_ZT = a_ZT / sqrt(tau_Z); // RL 24OCT2019: Combine taus for Z.
+ beta_ZY = a_ZY / sqrt(tau_Z);
+
+ // beta_XT = a_XT * tau_X; // RL 08DEC2019: Combine taus for X.
+ // beta_XY = a_XY * tau_X;
+ // beta_ZT = a_ZT * tau_Z; // RL 24OCT2019: Combine taus for Z.
+ // beta_ZY = a_ZY * tau_Z;
- alpha_T = sigma_T .* alpha_T_raw;
+ alpha_T = sigma_T * alpha_T_raw;
alpha_Y = sigma_Y * alpha_Y_raw;
}
@@ -65,11 +69,13 @@ model {
Z_obs ~ normal(0, 1);
Z_cens ~ normal(0, 1);
- // Intercepts
+ // Variance parameter for the intercepts
sigma_T ~ normal(0, sigma_tau);
sigma_Y ~ normal(0, sigma_tau);
+ // Intercepts
alpha_T_raw ~ normal(0, 1);
+ // alpha_T ~ logistic(0, 1);
alpha_Y_raw ~ normal(0, 1);
// According to
@@ -90,11 +96,11 @@ model {
a_ZY ~ normal(0, 1);
// nu = 6 -> nu/2 = 3
- tau_XT ~ gamma(3, 3);
- tau_XY ~ gamma(3, 3);
- tau_ZT ~ gamma(3, 3);
- tau_ZY ~ gamma(3, 3);
-
+ tau_X ~ gamma(3, 3);
+ tau_Z ~ gamma(3, 3);
+ // tau_X ~ normal(0, 1);
+ // tau_Z ~ normal(0, 1);
+
// Compute the regressions for the observed observations
for(i in 1:N_obs){
dec_obs[i] ~ bernoulli_logit(alpha_T[jj_obs[i]] + X_obs[i] * beta_XT + beta_ZT * Z_obs[i]);
@@ -107,9 +113,10 @@ model {
}
generated quantities {
+ // Array for predictions
int<lower=0, upper=1> y_est[N_cens];
-
- // Generate a draw from the posterior predictive.
+
+ // Generate a draw from the posterior predictive for the outcome.
for(i in 1:N_cens){
y_est[i] = bernoulli_logit_rng(alpha_Y + X_cens[i] * beta_XY + beta_ZY * Z_cens[i]);
}
--
GitLab