Final models

Author: JohnnyBeet

basic_fit.stan

@@ -8,6 +8,7 @@ data{
 
 parameters{
     real distance_coeff;
+    real mean;
     vector[4] traffic_level_coeff;
     real meal_prep_coeff;
     real<lower=0> sigma;
@@ -16,33 +17,28 @@ parameters{
 transformed parameters {
     vector[N] mu;
     for(i in 1:N){
-        mu[i] = exp(distance_coeff * distance[i] + traffic_level_coeff[traffic_level[i]] + meal_prep_coeff * meal_preparation_time[i]+3);
+        mu[i] = exp(distance_coeff * distance[i] + traffic_level_coeff[traffic_level[i]] + meal_prep_coeff * meal_preparation_time[i]+mean);
     }
 
 }
 
 model{
-    distance_coeff ~ normal(0,0.3);
-    meal_prep_coeff ~ normal(0,0.3);
+    mean ~ normal(3, 0.1);
+    distance_coeff ~ normal(0,0.5);
+    meal_prep_coeff ~ normal(0,0.5);
     sigma ~ exponential(0.5);
     delivery_times ~ inv_gamma(pow(mu, 2) / pow(sigma, 2) + 2, pow(mu,3) / pow(sigma, 2) + mu);
-    // delivery_times ~ normal(mu, sigma);
-    traffic_level_coeff[1] ~ normal(0, 0.3);
-    traffic_level_coeff[2] ~ normal(0, 0.3);
-    traffic_level_coeff[3] ~ normal(0, 0.3);
-    traffic_level_coeff[4] ~ normal(0, 0.3);
+    traffic_level_coeff[1] ~ normal(0, 0.5);
+    traffic_level_coeff[2] ~ normal(0, 0.5);
+    traffic_level_coeff[3] ~ normal(0, 0.5);
+    traffic_level_coeff[4] ~ normal(0, 0.5);
 }
 
 generated quantities {
     vector[N] delivery_time;
-    vector[N] exp_argument;
-    vector[N] exp_argument_withMeal;
     vector[N] log_lik;
     for (i in 1:N){
         delivery_time[i] = inv_gamma_rng(pow(mu[i], 2) / pow(sigma, 2) + 2, pow(mu[i],3) / pow(sigma, 2) + mu[i]);
         log_lik[i] = inv_gamma_lpdf(delivery_times[i] | pow(mu[i], 2) / pow(sigma, 2) + 2, pow(mu[i],3) / pow(sigma, 2) + mu[i]);
-        // delivery_time[i] = normal_rng(mu[i], sigma);
-        exp_argument_withMeal[i] = distance_coeff * distance[i] + traffic_level_coeff[traffic_level[i]] + meal_prep_coeff * meal_preparation_time[i]+3;
-        // exp_argument[i] = distance_coeff * distance[i] + traffic_level_coeff[traffic_level[i]];
     }
 }

basic_ppc.stan

@@ -7,31 +7,21 @@ data{
 
 generated quantities {
    real distance_coeff = normal_rng(0, 0.3);
+   real meal_prep_coeff = normal_rng(0, 0.3);
+   real mean = normal_rng(3, 0.1);
+
    array[4] real traffic_level_coeff;
    traffic_level_coeff[1] = normal_rng(0, 0.3);
    traffic_level_coeff[2] = normal_rng(0, 0.3);
    traffic_level_coeff[3] = normal_rng(0, 0.3);
    traffic_level_coeff[4] = normal_rng(0, 0.3);
-   // traffic_level_coeff[1] = 0;
-   // traffic_level_coeff[2] = 0;
-   // traffic_level_coeff[3] = 0;
-   // traffic_level_coeff[4] = 0;
-   real meal_prep_coeff = normal_rng(0, 0.3);
 
    array[N] real mu;
    real sigma = exponential_rng(0.5);
    array[N] real delivery_times;
-   array[N] real temp1;
-   array[N] real gamma_alpha;
-   array[N] real gamma_beta;
-   array[N] real temp2;
 
    for (i in 1:N){
-      temp1[i] = distance_coeff*distance[i] + traffic_level_coeff[traffic_level[i]]+meal_prep_coeff*meal_preparation_time[i]+3;
-      // temp2[i] = exp(distance_coeff*distance[i] + traffic_level_coeff[traffic_level[i]]);
-      mu[i] = exp(distance_coeff*distance[i] + traffic_level_coeff[traffic_level[i]]+meal_prep_coeff*meal_preparation_time[i]+3);
-      gamma_alpha[i] = pow(mu[i], 2) / pow(sigma, 2) + 2;
-      gamma_beta[i] = pow(mu[i],3) / pow(sigma, 2) + mu[i];
+      mu[i] = exp(distance_coeff*distance[i] + traffic_level_coeff[traffic_level[i]]+meal_prep_coeff*meal_preparation_time[i]+mean);
       delivery_times[i] = inv_gamma_rng(pow(mu[i], 2) / pow(sigma, 2) + 2, pow(mu[i],3) / pow(sigma, 2) + mu[i]);
    }
 }

distribution_from_data.png

@@ -5,11 +5,12 @@ data{
     vector[N] delivery_times;       // Delivery times
     array[N] int traffic_level;           // Traffic levels
     array[N] real delivery_person_rating;
-    array[N] int number_of_deliveiers;
+    array[N] int number_of_deliveries;
 }
 
 parameters{
     real distance_coeff;
+    real mean;
     vector[4] traffic_level_coeff;
     real meal_prep_coeff;
     real<lower=0> sigma;
@@ -20,18 +21,18 @@ parameters{
 transformed parameters {
     vector[N] mu;
     for(i in 1:N){
-        mu[i] = exp(distance_coeff * distance[i] + traffic_level_coeff[traffic_level[i]] + meal_prep_coeff * meal_preparation_time[i]+person_rating_coeff*delivery_person_rating[i]+deliveries_number_coeff[number_of_deliveiers[i]]+3);
+        mu[i] = exp(distance_coeff * distance[i] + traffic_level_coeff[traffic_level[i]] + meal_prep_coeff * meal_preparation_time[i]+person_rating_coeff*delivery_person_rating[i]+deliveries_number_coeff[number_of_deliveries[i]]+mean);
     }
 
 }
 
 model{
+    mean ~ normal(3, 0.1);
     person_rating_coeff ~ normal(0,0.3);
     distance_coeff ~ normal(0,0.3);
     meal_prep_coeff ~ normal(0,0.3);
     sigma ~ exponential(0.5);
     delivery_times ~ inv_gamma(pow(mu, 2) / pow(sigma, 2) + 2, pow(mu,3) / pow(sigma, 2) + mu);
-    // delivery_times ~ normal(mu, sigma);
     traffic_level_coeff[1] ~ normal(0, 0.3);
     traffic_level_coeff[2] ~ normal(0, 0.3);
     traffic_level_coeff[3] ~ normal(0, 0.3);
@@ -44,14 +45,9 @@ model{
 
 generated quantities {
     vector[N] delivery_time;
-    vector[N] exp_argument;
-    vector[N] exp_argument_withMeal;
     vector[N] log_lik;
     for (i in 1:N){
         delivery_time[i] = inv_gamma_rng(pow(mu[i], 2) / pow(sigma, 2) + 2, pow(mu[i],3) / pow(sigma, 2) + mu[i]);
         log_lik[i] = inv_gamma_lpdf(delivery_times[i] | pow(mu[i], 2) / pow(sigma, 2) + 2, pow(mu[i],3) / pow(sigma, 2) + mu[i]);
-        // delivery_time[i] = normal_rng(mu[i], sigma);
-        exp_argument_withMeal[i] = distance_coeff * distance[i] + traffic_level_coeff[traffic_level[i]] + meal_prep_coeff * meal_preparation_time[i]+person_rating_coeff*delivery_person_rating[i]+deliveries_number_coeff[number_of_deliveiers[i]]+3;
-        // exp_argument[i] = distance_coeff * distance[i] + traffic_level_coeff[traffic_level[i]];
     }
 }