More documentation.

fastlib/trunk/contrib/dongryel/thesis_research/core/monte_carlo/mean_variance_pair.h

@@ -1,4 +1,7 @@
 /** @file mean_variance_pair.h
+ *
+ *  The class implementation that represents a running sample mean and
+ *  variance pair.
  *
  *  @author Dongryeol Lee (dongryel@cc.gatech.edu)
  */
@@ -8,47 +11,79 @@
 
 namespace core {
 namespace monte_carlo {
+
+/** @brief The mean variance pair class.
+ */
 class MeanVariancePair {
 
   private:
+
+    /** @brief The number of samples gathered.
+     */
     int num_samples_;
 
+    /** @brief The sample mean.
+     */
     double sample_mean_;
 
+    /** @brief The sample variance.
+     */
     double sample_variance_;
 
   public:
 
+    /** @brief The default constructor that sets everything to zero.
+     */
     MeanVariancePair() {
       SetZero();
     }
 
+    /** @brief The copy constructor.
+     */
     MeanVariancePair(const MeanVariancePair &pair_in) {
       CopyValues(pair_in);
     }
 
+    /** @brief Copies another MeanVariancePair object.
+     */
     void CopyValues(const MeanVariancePair &pair_in) {
       num_samples_ = pair_in.num_samples();
       sample_mean_ = pair_in.sample_mean();
       sample_variance_ = pair_in.sample_variance();
     }
 
+    /** @brief The number of samples gathered is returned.
+     */
     int num_samples() const {
       return num_samples_;
     }
 
+    /** @brief Returns the sample mean.
+     */
     double sample_mean() const {
       return sample_mean_;
     }
 
+    /** @brief Returns the variance of the sample mean.
+     */
     double sample_mean_variance() const {
-      return sample_variance_ / ((double) num_samples_);
+      return sample_variance_ / static_cast<double>(num_samples_);
     }
 
+    /** @brief Returns the sample variance.
+     */
     double sample_variance() const {
-      return sample_variance_;
+
+      // Note that this function scales this way to return the proper
+      // variance.
+      return sample_variance_ * (
+               static_cast<double>(num_samples_) /
+               static_cast<double>(num_samples_ - 1));
     }
 
+    /** @brief Returns a scaled interval centered around the current
+     *         sample mean with the given standard deviation factor.
+     */
     void scaled_interval(
       double scale_in, double standard_deviation_factor,
       core::math::Range *interval_out) const {
@@ -68,17 +103,16 @@ class MeanVariancePair {
       interval_out->hi = scale_in * (sample_mean_ + error);
     }
 
+    /** @brief Sets everything to zero.
+     */
     void SetZero() {
       num_samples_ = 0;
       sample_mean_ = 0;
       sample_variance_ = 0;
     }
 
-    void Add(const MeanVariancePair &mv_pair_in) {
-      sample_mean_ += mv_pair_in.sample_mean();
-      sample_variance_ += mv_pair_in.sample_variance();
-    }
-
+    /** @brief Pushes a sample in.
+     */
     void push_back(double sample) {
 
       // Update the number of samples.
@@ -94,7 +128,7 @@ class MeanVariancePair {
                          ((double) num_samples_);
     }
 };
-};
-};
+}
+}
 
 #endif

fastlib/trunk/contrib/dongryel/thesis_research/mlpack/mixed_logit_dcm/mixed_logit_dcm_dev.h

@@ -247,6 +247,27 @@ double MixedLogitDCM<TableType>::IntegrationSampleError_(
   // Assumption: num_active_people in both samples are equal.
   double simulation_error = 0;
 
+  // Loop over each active people.
+  for(int i = 0; i < first_sample.num_active_people(); i++) {
+
+    // Get the active person index.
+    int person_index = table_.shuffled_indices_for_person(i);
+
+    // Get the integration samples for both samples.
+    const std::vector< arma::vec > &first_integration_samples =
+      first_sample.integration_samples();
+    const std::vector< arma::vec > &second_integration_samples =
+      second_sample.integration_samples();
+
+    // First compute the average difference.
+    core::monte_carlo::MeanVariancePair difference;
+    for(unsigned j = 0; j < first_integration_samples.size(); j++) {
+      double difference = ;
+      average_difference += ;
+    }
+    simulation_error += ;
+  }
+
   // Lastly divide by squared of the number of active people.
   simulation_error /=
     core::math::Sqr(static_cast<double>(first_sample.num_active_people()));