Merge pull request #191 from andrjohns/feature/basic_est_update

Feature/basic est update

Author: rok-cesnovar

basic_estimators/bernoulli.stan

@@ -1,12 +1,11 @@
 data { 
   int<lower=0> N; 
-  int<lower=0,upper=1> y[N];
+  int<lower=0, upper=1> y[N];
 } 
 parameters {
-  real<lower=0,upper=1> theta;
+  real<lower=0, upper=1> theta;
 } 
 model {
-  theta ~ beta(1,1);
-  for (n in 1:N) 
-    y[n] ~ bernoulli(theta);
+  theta ~ beta(1, 1);
+  y ~ bernoulli(theta);
 }

basic_estimators/negative_binomial.stan

@@ -7,8 +7,7 @@ parameters {
   real<lower=0> beta;
 }
 model { 
-  alpha ~ cauchy(0,10);
-  beta ~ cauchy(0,10);
-  for (i in 1:N)
-    y[i] ~ neg_binomial(alpha, beta);
+  alpha ~ cauchy(0, 10);
+  beta ~ cauchy(0, 10);
+  y ~ neg_binomial(alpha, beta);
 }

basic_estimators/negative_binomial2.stan

@@ -4,13 +4,11 @@ data {
 }
 parameters {
   real<lower=0> alpha;
-  real<lower=0,upper=1> p_success;
+  real<lower=0, upper=1> p_success;
 }
 transformed parameters {
-   real<lower=0.0> beta;
-   beta <- p_success / (1.0 - p_success);
+   real<lower=0.0> beta = p_success / (1.0 - p_success);
 }
 model { 
-  for (i in 1:N)
-    y[i] ~ neg_binomial(alpha, beta);
+  y ~ neg_binomial(alpha, beta);
 }

basic_estimators/normal_censored.stan

@@ -10,8 +10,7 @@ parameters {
 model {
   for (n in 1:N_observed)
     y[n] ~ normal(mu,1.0) T[,U];
-  increment_log_prob(N_censored 
-                     * log1m(normal_cdf(U,mu,1.0)));
+  target += N_censored * normal_lccdf(U | mu, 1);
 }
 
 

basic_estimators/normal_loc.stan

@@ -1,15 +1,9 @@
 transformed data {
-    real y[5];
-    y[1] <- 2.0;
-    y[2] <- 1.0;
-    y[3] <- -0.5;
-    y[4] <- 3.0;
-    y[5] <- 0.25;
+  vector[5] y = [2, 1, -0.5, 3, 0.25]';
 }
 parameters {
-    real mu;
+  real mu;
 }
 model {
-    for (n in 1:5)
-        y[n] ~ normal(mu,1.0);
-}
+  y ~ normal(mu, 1);
+}

basic_estimators/normal_mixture.stan

@@ -3,17 +3,19 @@
 
 data {
   int<lower=0>  N;
-  real y[N];
+  vector[N] y;
 }
 parameters {
   real<lower=0,upper=1> theta;
-  real mu[2];
+  vector[2] mu;
 }
 model {
   theta ~ uniform(0,1); // equivalently, ~ beta(1,1);
-  for (k in 1:2)
-    mu[k] ~ normal(0,10);
-  for (n in 1:N)
-    target += log_mix(theta, normal_lpdf(y[n]|mu[1],1.0), normal_lpdf(y[n]|mu[2],1.0));
+  mu ~ normal(0,10);
+  for (n in 1:N) {
+    target += log_mix(theta,
+                      normal_lupdf(y[n] | mu[1], 1),
+                      normal_lupdf(y[n] | mu[2], 1));
+  }
 }
 

basic_estimators/normal_mixture_k.stan

@@ -1,20 +1,20 @@
 data {
   int<lower=1> K;
   int<lower=1> N;
-  real y[N];
+  vector[N] y;
 }
 parameters {
   simplex[K] theta;
-  real mu[K];
-  real<lower=0,upper=10> sigma[K];
+  vector[K] mu;
+  vector<lower=0, upper=10>[K] sigma;
 }
 model {
-  real ps[K];              
+  vector[K] ps[N];              
   mu ~ normal(0,10);
   for (n in 1:N) {
-    for (k in 1:K)
-      ps[k] <- log(theta[k]) 
-               + normal_log(y[n],mu[k],sigma[k]);
-    increment_log_prob(log_sum_exp(ps));    
+    for (k in 1:K) {
+      ps[n][k] = normal_lupdf(y[n] | mu[k], sigma[k]);
+    }
   }
+  target += log_mix(theta, ps);  
 }

basic_estimators/normal_mixture_k_prop.stan

@@ -1,37 +1,32 @@
 data {
   int<lower=1> K;
   int<lower=1> N;
-  real y[N];
+  vector[N] y;
 }
 parameters {
   simplex[K] theta;
   simplex[K] mu_prop;
   real mu_loc;
   real<lower=0> mu_scale;
-  real<lower=0> sigma[K];
+  vector<lower=0>[K] sigma;
 }
 transformed parameters {
-  ordered[K] mu;
-  mu <- mu_loc + mu_scale * cumulative_sum(mu_prop);
+  ordered[K] mu = mu_loc + mu_scale * cumulative_sum(mu_prop);
 }
 model {
+  vector[K] ps[N];
+
   // prior
-  mu_loc ~ cauchy(0,5);               
-  mu_scale ~ cauchy(0,5);
-  sigma ~ cauchy(0,5);
+  mu_loc ~ cauchy(0, 5);               
+  mu_scale ~ cauchy(0, 5);
+  sigma ~ cauchy(0, 5);
 
   // likelihood
-  { 
-    real ps[K];
-    vector[K] log_theta;
-    log_theta <- log(theta);
-
-    for (n in 1:N) {
-      for (k in 1:K) {
-        ps[k] <- log_theta[k]
-                 + normal_log(y[n],mu[k],sigma[k]);
-      }
-      increment_log_prob(log_sum_exp(ps));    
-    }
+  for (n in 1:N) {
+    for (k in 1:K) {
+      ps[n][k] = normal_lupdf(y[n] | mu[k], sigma[k]);
+    } 
   }
+
+  target += log_mix(theta, ps);
 }

basic_estimators/normal_truncated.stan

@@ -1,13 +1,13 @@
 data {
   real U;
   int<lower=1> N;
-  real<upper=U> y[N];
+  vector<upper=U>[N] y;
 }
 parameters {
   real mu;
-  real<lower=0,upper=2> sigma;
+  real<lower=0, upper=2> sigma;
 }
 model {
   for (n in 1:N)
-    y[n] ~ normal(mu,sigma) T[,U];
+    y[n] ~ normal(mu, sigma) T[,U];
 }