Make comments for each cells.

- 1. import necessary libraries \# \# import pandas as pd import numpy as np import random import matplotlib.pyplot as plt \%matplotlib inline random. seed(30) 2. data wrangling df= pd.read_csv('final_exam_dataset2.csv') df.info() top_k =10 df.head(top_k) df.groupby (['control_group', 'converted']).agg( 'count') df.drop(df.query("control_group == True and landing_page == 'new'" ).index, inplace=True) df.drop(df.query("control_group == False and landing_page == 'old'").index, inplace=True) df.groupby (['control_group', 'converted']).agg( 'count') df.info() 3. remove duplicates df[df.duplicated([ 'user_id'], keep=False) ] id user_id timestamp control_group landing_page converted df.drop_duplicates(['user_id'], inplace=True) assert len(df['user_id'].unique ())=df[ 'user_id'].size df.info() 3. basic statistics df[ ' converted' ].mean() df.groupby (['control_group']).describe() df.groupby (['control_group']).agg(\{'converted' : ['sum', 'count', 'mean'] } ) df[['control_group', 'converted']].groupby(['control_group'] ).agg( 'mean' ).T df. head() 4. Probablities and countings of old pages and new pages A/B Testing: Calculate z-score and p-value (area) Over Dataset import statsmodels.api as sm A/B Testing: Calculate z-score corresponding to alpha =0.05 from scipy.stats import norm print('p-value: ", norm. cdf(z_score)) \# Tells us how significant our z-score is \# Single-side test at 95\% confidence level, we calculate: print('z_alpha:', norm.ppf(1 - (0.05))) A/B Testing: Calculate Beta and Power corresponding to alpha =0.05, effect Size from statsmodels.stats import power as pwr from statsmodels.stats.proportion import proportion_effectsize effect_size = proportion_effectsize(probability_newpage, probability_oldpage) ratio = (newpage_counts / oldpage_counts) beta =1 - power print('power: ', power)