def main():
state = pd.read_spss('/Users/kellenbullock/Desktop/Geographic Analysis II/Data/5303_EX_A.sav')
Counties = state.query("Scale == 'Counties'")
Schools = state.query("Scale == 'Schools'")
Tracts = state.query("Scale == 'Tracts'")
assigned_var_c = Counties[['Pct_Black', 'Pct_Two_Plus', 'Pct_SNAP', 'Pct_FIRE_I', 'Pct_Poverty', 'Pct_Unemp', 'Med_HomeValue', 'Pct_White', 'Pct_BlueCollar_O', 'Pct_Hispanic']]
assigned_var_s = Schools[['Pct_Black', 'Pct_Two_Plus', 'Pct_SNAP', 'Pct_FIRE_I', 'Pct_Poverty', 'Pct_Unemp', 'Med_HomeValue', 'Pct_White', 'Pct_BlueCollar_O', 'Pct_Hispanic']]
assigned_var_t = Tracts[['Pct_Black', 'Pct_Two_Plus', 'Pct_SNAP', 'Pct_FIRE_I', 'Pct_Poverty', 'Pct_Unemp', 'Med_HomeValue', 'Pct_White', 'Pct_BlueCollar_O', 'Pct_Hispanic']]
assigned_var_t = assigned_var_t.reset_index()
assigned_var_s = assigned_var_s.reset_index()
assigned_var_s = assigned_var_s.drop(columns=['index'])
assigned_var_t = assigned_var_t.drop(columns=['index'])
# New Variables
var_c = Counties[['Pct_Unemp', 'Med_HomeValue', 'Pct_White', 'Pct_BlueCollar_O', 'Pct_Hispanic']]
var_s = Schools[['Pct_Unemp', 'Med_HomeValue', 'Pct_White', 'Pct_BlueCollar_O', 'Pct_Hispanic']]
var_t = Tracts[['Pct_Unemp', 'Med_HomeValue', 'Pct_White', 'Pct_BlueCollar_O', 'Pct_Hispanic']]
var_t = var_t.reset_index()
var_s = var_s.reset_index()
var_t = var_t.drop(columns=['index'])
var_s = var_s.drop(columns=['index'])
# 1.
#EDA.figures(var_c, 'Counties')
#EDA.figures(var_t, 'Tracts')
#EDA.figures(var_s, 'Schools')
print('=====County=======')
#EDA.Descrptives(Counties, var_c)
print('======== Tracts =========')
#EDA.Descrptives(Tracts, var_t)
print('====== School Districts =======')
#EDA.Descrptives(Schools, var_s)
# 1 part b:
#test_trans(assigned_var_c, 'Pct_Black')
print('************')
#test_trans(assigned_var_c, 'Pct_Hispanic')
# 2,
# This will make a pearson's r correlation matrix at the County scale
# These do not work in the spyder IDE. Please see the jupyter notebook for outputs.
corr = assigned_var_c.corr()
corr.style.background_gradient(cmap='coolwarm').set_precision(3)
EDA.df_to_pdf(corr, 'Matrix_1')
''' Easy way to order Correlations but without signs:
correlations = assigned_var_c.corr().abs()
stack = correlations.unstack()
stack_order = s.sort_values(kind='quicksort')
'''
# 2. a. 8 strongest Person's correaltions. There was no easy way to do this I had to pull everythong by hand.
corr_table = {
'Variables': ['Pct_Poverty / Pct_SNAP', 'Pct_Unemp / Pct_SNAP', 'Pct_White / Pct_Unemp', 'Pct_White / Pct_SNAP', 'Pct_Unemp / Pct_Poverty', 'Pct_White / Pct_Two_Plus', 'Pct_White / Pct_Poverty', 'Pct_Fire_I / Pct_BlueCollar_O'],
'Correlation': [0.757, 0.722, - 0.672, - 0.645, 0.605, - 0.599, - 0.593, - 0.574]
}
all_scale = {
'Variables': ['Pct_Poverty / Pct_SNAP', 'Pct_Unemp / Pct_SNAP', 'Pct_White / Pct_Unemp', 'Pct_White / Pct_SNAP', 'Pct_Unemp / Pct_Poverty', 'Pct_White / Pct_Two_Plus', 'Pct_White / Pct_Poverty', 'Pct_Fire_I / Pct_BlueCollar_O'],
'County': [0.757, 0.722, - 0.672, - 0.645, 0.605, - 0.599, - 0.593, - 0.574],
'Tract': [0.773, 0.679, -0.474, -0.564, 0.628, -0.296, -0.495, -0.584],
'School District': [0.732, 0.590, -0.440, -0.515, 0.494, -0.569, -0.442, -0.380]
}
pearson_r_table = pd.DataFrame(corr_table)
pearson_r_table.to_excel('County Correlation.xlsx')
EDA.df_to_pdf(pearson_r_table, 'Decending_corr')
all_scale = pd.DataFrame(all_scale)
all_scale.to_excel('Scale_Correlation.xlsx')
EDA.df_to_pdf(all_scale, 'Scale_Correlation')
# 3.
assigned_c_repub = Counties[['Pct_Repub', 'Pct_Black', 'Pct_Two_Plus', 'Pct_SNAP', 'Pct_FIRE_I', 'Pct_Poverty', 'Pct_Unemp', 'Med_HomeValue', 'Pct_White', 'Pct_BlueCollar_O', 'Pct_Hispanic']]
assigned_c_repub = assigned_c_repub.reset_index()
assigned_c_repub = assigned_c_repub.drop(columns=['index'])
assigned_c_repub['Pct_Black'] = Transformations.log_trans(assigned_c_repub, 'Pct_Black')
assigned_c_repub['Pct_Hispanic'] = Transformations.log_trans(assigned_c_repub, 'Pct_Hispanic')
corr2 = assigned_c_repub.corr()
corr2.style.background_gradient(cmap='coolwarm').set_precision(3)
EDA.df_to_pdf(corr2, 'Matrix_2')
# c.
chosen = corr2[['Pct_Repub', 'Pct_Poverty', 'Pct_Unemp', 'Pct_White', 'Pct_SNAP']]
x = ['Pct_Repub', 'Pct_Poverty', 'Pct_Unemp', 'Pct_White', 'Pct_SNAP']
y = 'Pct_Repub'
corr_scatter(chosen, x, y)
# c. 4 scatter plots with y axis being pct_republican
chosen = assigned_c_repub[['Pct_Repub', 'Pct_Poverty', 'Pct_Unemp', 'Pct_White', 'Pct_SNAP']]
# Running partial correlation 4 times:
Partial_Corr.partial_corr(chosen[['Pct_Repub', 'Pct_Poverty']])
Partial_Corr.partial_corr(chosen[['Pct_Repub', 'Pct_White']])
Partial_Corr.partial_corr(chosen[['Pct_Repub', 'Pct_SNAP']])
Partial_Corr.partial_corr(chosen[['Pct_Repub', 'Pct_Unemp']])
# 5.
bivar_regres(chosen['Pct_Repub'], chosen['Pct_Poverty']) # Model 1
bivar_regres(chosen['Pct_Repub'], chosen['Pct_SNAP']) # Model 2
bivar_regres(chosen['Pct_Repub'], chosen['Pct_White']) # Model 3
bivar_regres(chosen['Pct_Repub'], chosen['Pct_Unemp']) # Model 4
