# -*- coding: utf-8 -*-
"""
Created on Wed Dec 21 10:53:12 2022

@author: janmo
"""
import statistics
from scipy.stats import mannwhitneyu
import pandas as pd
import numpy as np
import random
import math
import seaborn as sns
import matplotlib.pyplot as plt
import statsmodels.api as sm

from scipy.stats import wilcoxon
from sklearn.linear_model import LogisticRegression

# Read in and select Data
Data_A = pd.read_csv('Final/all_apps_wide-2022-12-20 (1).csv')
Data_A = Data_A.iloc[2:,:]
Data_A = Data_A.loc[(pd.isna(Data_A['participant._current_app_name'])==False) & (pd.isna(Data_A['lottery_risk_game.1.player.BDM'])==False), :]
Data_A = Data_A.loc[Data_A['participant.id_in_session']!=215]
Data_A['treatment'] = Data_A.apply(lambda row: 0 if row['lottery_risk_game.1.player.treatment']=='baseline' else 1, axis=1)


# create list with plane feature names
def cut_feature_string(string):
    string_cut = string[27:]
    return string_cut

survey_features = Data_A.columns[55:68]
survey_features_cut = list(map(cut_feature_string, survey_features))

for i, column_name in zip(range(55,68), survey_features_cut):
    
    fig, axes = plt.subplots(figsize=(16,8))
    sns.histplot(Data_A.iloc[:,i], bins=10, kde=True, stat="density", linewidth=0).set_title(column_name, fontsize=20, weight='bold')
    axes.set_xlabel(column_name, fontsize=15, weight='bold')
    axes.set_ylabel('Density', fontsize=15, weight='bold')
    plt.show()
    
    print(column_name)
    print(Data_A.iloc[:,i].describe())
    print('- - -')

### Aggregate construcs
Data_A['Cognitive_trust'] = (Data_A['lottery_risk_game.1.player.algo_expert'] + Data_A['lottery_risk_game.1.player.algo_knowledge'])/2
Data_A['Emotional_trust'] = (Data_A['lottery_risk_game.1.player.feel_secure'] + Data_A['lottery_risk_game.1.player.feel_comfortable'] + Data_A['lottery_risk_game.1.player.feel_content'])/3
Data_A['Privacy']         = (Data_A['lottery_risk_game.1.player.privacy_1'] + Data_A['lottery_risk_game.1.player.privacy_2'])/2


# Select treatment participants
Data_A_treatment = Data_A.loc[Data_A['lottery_risk_game.1.player.treatment']=='treatment']


# Variable list
var_list = ['lottery_risk_game.1.player.BDM',
            'lottery_risk_game.1.player.perceived_accuracy',
            'lottery_risk_game.1.player.perceived_rmse',
            'Privacy',
            'Cognitive_trust',
            'Emotional_trust',
            'lottery_risk_game.1.player.transparency',
            'lottery_risk_game.1.player.power']

# Mean ans stdev of variables - Baseline and Treatment 
for var in var_list:
    print('# - - - - - -', var, '- - - - - - - #')
    print('Means:')
    print(Data_A.groupby('lottery_risk_game.1.player.treatment')[var].mean())
    print(' ')
    print('Standard deviations:')
    print(Data_A.groupby('lottery_risk_game.1.player.treatment')[var].std())
    print(' ')
    
    print('OLS:')
    X = Data_A['treatment']
    y = Data_A[var]
    X = sm.add_constant(X)
    X.head()
    h1_est_1 = sm.OLS(y, X).fit()
    print(h1_est_1.summary())
    print("\n", "\n")


    # Whitneyutest
    X_baseline  = Data_A.loc[Data_A['treatment']==0, var]
    y_treatment = Data_A.loc[Data_A['treatment']==1, var]
    
    print(mannwhitneyu(X_baseline, y_treatment))
    print('________________________________________________',"\n", "\n", "\n", "\n")

# Mean ans stdev of variables - Baseline and Treatment min. x

# Prepare treatment participants
# Check which pars did not choose features with threshold
Data_A['conducted_dec_fs?']=0
dummy_metalist = []

for i in range(0,len(Data_A)):
    dummy_list = []
    
    for j in range(68,81):
        dummy = Data_A.iloc[i,j]
        dummy_list.append(dummy)
        
    if sum(dummy_list) >= 1:
        check_fs = True
    else:
        check_fs = False
        
    if check_fs==False:
        Data_A['conducted_dec_fs?'].iloc[i]=0 #0 if no dec fs conducted
    else:
        Data_A['conducted_dec_fs?'].iloc[i]=1
        
    dummy_metalist.append(dummy_list)
    
print(len(Data_A.loc[Data_A['conducted_dec_fs?']==1])/len(Data_A_treatment))


var_list = ['lottery_risk_game.1.player.BDM',
            'lottery_risk_game.1.player.perceived_accuracy',
            'lottery_risk_game.1.player.perceived_rmse',
            'Privacy',
            'Cognitive_trust',
            'Emotional_trust',
            'lottery_risk_game.1.player.transparency',
            'lottery_risk_game.1.player.power']

Data_A_min = Data_A.loc[(Data_A['lottery_risk_game.1.player.treatment']=='baseline') | ((Data_A['lottery_risk_game.1.player.treatment']=='treatment') & (Data_A['conducted_dec_fs?']==1))]
for var in var_list:
    print('# - - - - - -', var, '- - - - - - - #')
    print('Means:')
    print(Data_A_min.groupby('treatment')[var].mean())
    print(' ')
    print('Standard deviations:')
    print(Data_A_min.groupby('treatment')[var].std())
    print(' ')
    
    print('OLS:')
    X = Data_A_min['treatment']
    y = Data_A_min[var]
    X = sm.add_constant(X)
    X.head()
    h1_est_1 = sm.OLS(y, X).fit()
    print(h1_est_1.summary())
    print('________________________________________________',"\n", "\n", "\n", "\n")

    # Whitneyutest
    X_baseline  = Data_A.loc[Data_A['treatment']==0, var]
    y_treatment = Data_A.loc[Data_A['treatment']==1, var]
    
    print(mannwhitneyu(X_baseline, y_treatment))
    print('________________________________________________',"\n", "\n", "\n", "\n")

print(Data_A_min['conducted_dec_fs?'].sum(), 'participants involved')


### Median of # hidden features
sum_hidden_features_list = []

for i in dummy_metalist:
    sum_hidden_features = sum(i)
    sum_hidden_features_list.append(sum_hidden_features)

for quantile in [0.5, 0.6, 0.7, 0.8, 0.9]:
    print(quantile, 'quantile:', np.quantile(sum_hidden_features_list, quantile))




# Visualization BDM for each treatment threshold
BDM_mean_per_threshold = []
for threshold in range(0,13):
    print(threshold)

    Data_A['conducted_dec_fs?']=0
    for i in range(0,len(Data_A)):
        dummy_list = []
        
        for j in range(68,81):
            dummy = Data_A.iloc[i,j]
            dummy_list.append(dummy)
            
        if sum(dummy_list) >= threshold:
            check_fs = True
        else:
            check_fs = False
            
        if check_fs==False:
            Data_A['conducted_dec_fs?'].iloc[i]=0 #0 if no dec fs conducted
        else:
            Data_A['conducted_dec_fs?'].iloc[i]=1
    
    Data_A_min = Data_A.loc[(Data_A['lottery_risk_game.1.player.treatment']=='treatment') & (Data_A['conducted_dec_fs?']==1)]
    BDM_value = Data_A_min['lottery_risk_game.1.player.BDM'].mean()
    BDM_mean_per_threshold.append(BDM_value)
    print(threshold)

fig, axes = plt.subplots(figsize=(12,5))
sns.lineplot(list(range(1,14)), BDM_mean_per_threshold)
axes.set_xlabel('# minimum hidden features', fontsize=15, weight='bold')
axes.set_ylabel('Schadensrate', fontsize=15, weight='bold')
axes.set_title('Treatment WTP for different # of hidden features', fontsize=15, weight='bold')





# Backup 

# - - - WTP - - - #
X_baseline_test1  = Data_A.loc[Data_A['treatment']==0, 'lottery_risk_game.1.player.BDM'].isna().sum()
y_treatment_test1 = Data_A.loc[Data_A['treatment']==1, 'lottery_risk_game.1.player.BDM']

X_baseline_test2  = Data_A.loc[Data_A['treatment']==0, 'lottery_risk_game.1.player.perceived_accuracy'].isna().sum()
y_treatment_test2 = Data_A.loc[Data_A['treatment']==1, 'lottery_risk_game.1.player.perceived_accuracy'].isna().sum()

X_baseline_test2  = Data_A.loc[Data_A['treatment']==0, 'lottery_risk_game.1.player.perceived_rmse'].isna().sum()
y_treatment_test2 = Data_A.loc[Data_A['treatment']==1, 'lottery_risk_game.1.player.perceived_rmse'].isna().sum()

X_baseline_test2  = Data_A.loc[(Data_A['treatment']==0) & (Data_A['Privacy'].isna()==True), 'Privacy']
y_treatment_test2 = Data_A.loc[Data_A['treatment']==1, 'Privacy'].isna().sum()

mannwhitneyu(X_baseline_test2, y_treatment_test2)


print(Data_A.groupby('lottery_risk_game.1.player.treatment')['lottery_risk_game.1.player.BDM'].mean())
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['lottery_risk_game.1.player.BDM'].std())

X = Data_A['treatment']
y = Data_A['lottery_risk_game.1.player.BDM']
X = sm.add_constant(X)
X.head()
h1_est_1 = sm.OLS(y, X).fit()
print(h1_est_1.summary())

# - - - Perc_acc - - - #
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['lottery_risk_game.1.player.perceived_accuracy'].mean())
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['lottery_risk_game.1.player.perceived_accuracy'].std())

X = Data_A['treatment']
y = Data_A['lottery_risk_game.1.player.perceived_accuracy']
X = sm.add_constant(X)
X.head()
h1_est_1 = sm.OLS(y, X).fit()
print(h1_est_1.summary())

# - - - Perceived_rmse - - - #
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['lottery_risk_game.1.player.perceived_rmse'].mean())
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['lottery_risk_game.1.player.perceived_rmse'].std())

X = Data_A['treatment']
y = Data_A['lottery_risk_game.1.player.perceived_rmse']
X = sm.add_constant(X)
X.head()
h1_est_1 = sm.OLS(y, X).fit()
print(h1_est_1.summary())

# - - - Privacy - - - #
X = Data_A['treatment']
y = Data_A['lottery_risk_game.1.privacy']
X = sm.add_constant(X)
X.head()
h1_est_1 = sm.OLS(y, X).fit()
print(h1_est_1.summary())

# - - - Perceived_rmse - - - #
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['Privacy'].mean())
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['Privacy'].std())

X = Data_A['treatment']
y = Data_A['Privacy']
X = sm.add_constant(X)
X.head()
h1_est_1 = sm.OLS(y, X).fit()
print(h1_est_1.summary())

# - - - Cognitive trust - - - #
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['Cognitive_trust'].mean())
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['Cognitive_trust'].std())

X = Data_A['treatment']
y = Data_A['Cognitive_trust']
X = sm.add_constant(X)
X.head()
h1_est_1 = sm.OLS(y, X).fit()
print(h1_est_1.summary())

# - - - Emotional trust - - - #
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['Emotional_trust'].mean())
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['Emotional_trust'].std())

X = Data_A['treatment']
y = Data_A['Emotional_trust']
X = sm.add_constant(X)
X.head()
h1_est_1 = sm.OLS(y, X).fit()
print(h1_est_1.summary())

# - - - Emotional trust - - - #
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['Emotional_trust'].mean())
print(Data_A.groupby('lottery_risk_game.1.player.treatment')['Emotional_trust'].std())

X = Data_A['treatment']
y = Data_A['Emotional_trust']
X = sm.add_constant(X)
X.head()
h1_est_1 = sm.OLS(y, X).fit()
print(h1_est_1.summary())