#coefficients
c_range = list(df1.columns[25:29])
#sectors/ind
i_range = list(['oil & gas','mining','metals',
                'chemicals','telecommunications','utilities',
                'transportation','agribusiness','retail trade','general manufcaturing'])
#date range
d_range = list(df1['bs_rep_date_corr'][df1['bs_rep_date_corr'].dt.year>=2008].unique())

N = range (0,10000) #итерации
p = 0.05 #уровень доверия

res_store = pd.DataFrame(columns=['coefficient', 'sector', 'date', 'mi low', 'mi', 'mi upp','me'])

for c in c_range: #for each coefficient
    tmp = df1[['bs_rep_date_corr', 'ind_mapping.Ind_corr2', c]].copy().reset_index(drop=True)
    for i in i_range: #for each ind/sector
        for d in d_range: #for each date
            tmp1 = tmp.loc[(tmp['ind_mapping.Ind_corr2']==i) & 
                          (tmp['bs_rep_date_corr']==d) & 
                          (tmp[c]>=-1) & (tmp[c]<=1)].reset_index(drop=True)
            
            estimate_values = [] #empty df for saving iterations
            print(c,i,d)
            
            for n in N: #boot iterations
                tmp_rnd = tmp1.sample(frac=1.,replace=True)
                estimate_values.append(tmp_rnd.mean())
            estimate_values = pd.DataFrame(estimate_values)
                     
            mi = estimate_values.mean().values[0].round(5)
            mi_lowB = estimate_values.quantile(q=p/2).values[0].round(5)
            mi_uppB = estimate_values.quantile(q=1-p/2).values[0].round(5)
            me = estimate_values.median().values[0].round(5)
            
            #saving cycle result
            out = pd.DataFrame([[c, i, d, mi_lowB, mi, mi_uppB, me]], columns=res_store.columns)
            res_store = res_store.append(out)
print('done')