MedCohort.h

#pragma once
#ifndef __MED_COHORT_H__
#define __MED_COHORT_H__
 
//===================================================================================
//===================================================================================
 
 
#include <vector>
#include <SerializableObject/SerializableObject/SerializableObject.h>
#include <InfraMed/InfraMed/InfraMed.h>
#include <MedProcessTools/MedProcessTools/MedSamples.h>
 
using namespace std;
 
 
//===================================================================================
//===================================================================================
struct CohortRec : SerializableObject {
    int pid = -1;           
    int from = 0;           
    int to = 0;             
    int outcome_date = 0;   
    float outcome = -1;     
    string comments = "";   
 
    CohortRec() {};
    CohortRec(int _pid, int _from, int _to, int _outcome_date, float _outcome) {
        pid = _pid; from = _from; to = _to; outcome_date = _outcome_date; outcome = _outcome;
    }
 
    int init(map<string, string>& map);
 
    void get_string(string &to_str);
    int from_string(string &from_str);
};
 
//===================================================================================
//===================================================================================
struct SamplingParams : SerializableObject {
    float min_control_years = 0;    
    float max_control_years = 10;   
    float min_case_years = 0;   
    float max_case_years = 1;   
    int is_continous = 1;           
    int min_days_from_outcome = 30;     
    int jump_days = 180;    
    int min_year = 1900;    
    int max_year = 2100;    
    int gender_mask = 0x3;  
    int train_mask = 0x7;   
    int min_age = 0;    
    int max_age = 200;  
    string rep_fname;   
 
    vector<string> stick_to_sigs;       
    int take_closest = 0;   
    int take_all = 0; 
    int max_samples_per_id = (int)1e9; 
    string max_samples_per_id_method = "last"; 
 
    int init(map<string, string>& map);
};
 
//===================================================================================
//===================================================================================
struct IncidenceParams : SerializableObject {
    int from_year = 2007;   
    int to_year = 2013; 
    int start_date = 101; 
    int from_age = 30;  
    int to_age = 90;    
    int age_bin = 5;    
    int incidence_years_window = 1; 
    int incidence_days_win = -1;  
    int min_samples_in_bin = 20; 
    int gender_mask = 0x3;  
    int train_mask = 0x7;   
    string rep_fname;   
 
 
    int init(map<string, string>& map);
};
 
//===================================================================================
//===================================================================================
class MedCohort : SerializableObject {
 
 public:
    vector<CohortRec> recs; 
 
    void insert(int pid, int from, int to, int outcome_date, float outcome) { recs.push_back(CohortRec(pid, from, to, outcome_date, outcome)); }
 
    int read_from_file(string fname);
    int write_to_file(string fname);
 
    int read_from_bin_file(string fname) { return SerializableObject::read_from_file(fname); }
    int write_to_bin_file(string fname) { return SerializableObject::write_to_file(fname); }
 
    void get_pids(vector<int> &pids);
 
    //int print_general_stats();
 
    int create_incidence_file(IncidenceParams &i_params, string out_file, const string &debug_file = "");
 
    int create_sampling_file(SamplingParams &s_params, string out_sample_file);
    int create_samples(MedRepository& rep, SamplingParams &s_params, MedSamples& samples);
 
    int create_sampling_file_sticked(SamplingParams &s_params, string out_sample_file);
    int create_samples_sticked(MedRepository& rep, SamplingParams &s_params, MedSamples& samples);
};
 
 
//===================================================================================
// A few more MedSamples Helpers
 
// Scanner ::
// Given a MedSamples file , allows defining a sub-sample of it (can be all), 
// And define a list of tests and a list of base_tests
//
// The Scanner then allows the following:
// (1) Count for every test / base_test how many had at least N tests in a window W.
//     This helps in considering only variables that HAVE data.
// (2) Train a model M for each of:
//     - base_tests only
//     - base_tests + single test 
//       -> for all train/test group
//       -> only for the subgroup of points that HAS no missing values (and compare to the base just on those)
//
 
 
#endif