count_users_group = df_clear.pivot_table(index = 'group', values = 'user_id', aggfunc = ['nunique']).reset_index()

count_users_group.columns = ['group', 'n_user']

count_users_group

events_group_users = df_clear.pivot_table(index = 'events', columns = 'group', values = 'user_id',\

                                          aggfunc = 'nunique')\

                             .sort_values(by = 'A1', ascending = False)\

                             .reset_index().drop(4)

events_group_users['count_users'] =  events_group_users['A1'] + events_group_users['A2'] + events_group_users['B']

events_group_users['sum_A1_A2'] = events_group_users['A1'] + events_group_users['A2']

events_group_users['proportionA1'] = events_group_users['A1'] / count_users_group.iloc[0,1] * 100

events_group_users['proportionA2'] = events_group_users['A2'] / count_users_group.iloc[1,1] * 100

events_group_users['proportionB'] = events_group_users['B'] / count_users_group.iloc[2,1] * 100

events_group_users['proportionA1_A2'] = ((events_group_users['A1'] + events_group_users['A2'])\

                                        / (count_users_group.iloc[0,1] + count_users_group.iloc[1,1])) * 100

 

#создадим функцию

def test(successes1, successes2, trials1, trials2):  

    alpha = .05

    p1 = successes1/ trials1

    p2 = successes2 / trials2

    p_combined = (successes1 + successes2) / (trials1 + trials2)

    difference = p1 - p2

   

    z_value = difference / mth.sqrt(p_combined * (1 - p_combined) * (1/trials1 + 1/trials2))

    distr = stats.norm(0, 1)

    p_value = (1 - distr.cdf(abs(z_value))) * 2

   

    print('Событие:', event_name)

    print('p-значение: ', p_value)

 

    if p_value < alpha:

        print('Отвергаем нулевую гипотезу: между долями есть значимая разница')

    else:

        print('Не получилось отвергнуть нулевую гипотезу, нет оснований считать доли разными')

 

# теперь проверим группы A1 и A2 по всем событиям:

#for index, row in events_group_users.iterrows():

#    test(row["A1"], row["A2"], count_users_group['n_user'][0],count_users_group['n_user'][1])

#    print()

events_group_users['events'] = events_group_users.apply(lambda row: test(row['A1'], row['A2'], count_users_group['n_user'][0], count_users_group['n_user'][1]), axis=1)