Voodoo - Business Case - Pierre Bollack

# Base # ----------------------------------- import numpy as np import pandas as pd import seaborn as sns import matplotlib.pyplot as plt # Hypothesis Testing # ----------------------------------- import scipy.stats as stats

#import the dataset df = pd.read_csv('/work/Voodoo_Test_Business_Case.csv', delimiter = ';')

#check the dataset df.columns

df.info() # We can observe that the dataset is composed of 372834 observations and 18 features. # Moreover, there is no missing data.

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 372834 entries, 0 to 372833
Data columns (total 18 columns):
 #   Column              Non-Null Count   Dtype  
---  ------              --------------   -----  
 0   session_id          372834 non-null  object 
 1   platform            372834 non-null  object 
 2   open_at             372834 non-null  object 
 3   user_id             372834 non-null  object 
 4   country             372832 non-null  object 
 5   ab_cohort_id        372834 non-null  int64  
 6   segment_id          372834 non-null  int64  
 7   session_length      372834 non-null  int64  
 8   session_number_     372834 non-null  int64  
 9   publisher_revenue   372834 non-null  float64
 10  fs_shown            372834 non-null  int64  
 11  rv_shown            372834 non-null  int64  
 12  game_count          372834 non-null  int64  
 13  offline_game_count  372834 non-null  int64  
 14  acquired_at         372834 non-null  object 
 15  segment_name        372834 non-null  object 
 16  ab_name             372834 non-null  object 
 17  ab_cohort_name      372834 non-null  object 
dtypes: float64(1), int64(8), object(9)
memory usage: 51.2+ MB

df.head()

	session_idobject	platformobject	open_atobject	user_idobject	countryobject	ab_cohort_idint64	segment_idint64	session_lengthint64	session_number_int64	publisher_revenuefloat64	fs_shownint64	rv_shownint64	game_countint64	offline_game_countint64	acquired_atobject	segment_nameobject	ab_nameobject	ab_cohort_nameobject
0	e058d9e3-ea0c-4d16-a36e-3e9ed3840457	android	2020-02-09 11:27:45.717000	3675ed02-ea35-4f01-8375-2ee48b06ac5c	TN	790	783	36	5	0.00001	0	0	1	0	2020-02-08 10:35:38.630000	android_new_users	ab-test_android	gameTune
1	6b9d2f3d-fe82-45fb-9c67-9408934393b1	android	2020-02-09 11:04:38.491000	3675ed02-ea35-4f01-8375-2ee48b06ac5c	TN	790	783	158	4	0.00011	1	0	1	0	2020-02-08 10:35:38.630000	android_new_users	ab-test_android	gameTune
2	bdf0bb00-9241-4e2a-969b-c55267befa76	android	2020-02-09 10:59:48.496000	3675ed02-ea35-4f01-8375-2ee48b06ac5c	TN	790	783	90	3	0.00002	1	0	0	0	2020-02-08 10:35:38.630000	android_new_users	ab-test_android	gameTune
3	79ea0b6c-ea19-480d-b6b4-60acbef70241	android	2020-02-05 12:15:02.907000	51ea5d8d-9b28-4111-ab23-40961dc43321	AE	789	783	517	1	0.0555148	7	0	4	0	2020-02-05 12:15:02.907000	android_new_users	ab-test_android	xxHigh
4	f5f1f67f-ac44-45ba-9650-07b8081d897f	android	2020-02-07 06:03:13.014000	ecfbf8db-300f-4ceb-a533-47d9bf0c9fc7	AE	790	783	385	1	0.0142999999999999	2	0	5	0	2020-02-07 06:03:13.014000	android_new_users	ab-test_android	gameTune
5 rows × 18 columns

#Number of Unique User df.user_id.nunique()

# Summary Stats: publisher_revenue df.describe([0.01, 0.05, 0.10, 0.20, 0.80, 0.90, 0.95, 0.99])[["publisher_revenue"]].T

	countfloat64	meanfloat64	stdfloat64	minfloat64	1%float64	5%float64	10%float64	20%float64	50%float64	80%float64	90%float64	95%float64	99%float64	maxfloat64
publisher_revenue	372834	0.03766053830235974	0.05950961012931261	0	0	0	0.00001	0.0010395200000000069	0.0169956	0.060958359600000005	0.10096659200000002	0.14545459999999968	0.26984709199999973	2.891823927000002
1 rows × 14 columns

#histogram sns.distplot(df['publisher_revenue']);

/shared-libs/python3.7/py/lib/python3.7/site-packages/seaborn/distributions.py:2557: FutureWarning: `distplot` is a deprecated function and will be removed in a future version. Please adapt your code to use either `displot` (a figure-level function with similar flexibility) or `histplot` (an axes-level function for histograms).
  warnings.warn(msg, FutureWarning)

# Summary Stats: session_number_ df.describe([0.01, 0.05, 0.10, 0.20, 0.80, 0.90, 0.95, 0.99])[["session_number_"]].T

	countfloat64	meanfloat64	stdfloat64	minfloat64	1%float64	5%float64	10%float64	20%float64	50%float64	80%float64	90%float64	95%float64	99%float64	maxfloat64
session_number_	372834	4.73364821877833	5.541589133018086	0	1	1	1	1	3	7	11	15	27	137
1 rows × 14 columns

#histogram sns.distplot(df['session_number_']);

/shared-libs/python3.7/py/lib/python3.7/site-packages/seaborn/distributions.py:2557: FutureWarning: `distplot` is a deprecated function and will be removed in a future version. Please adapt your code to use either `displot` (a figure-level function with similar flexibility) or `histplot` (an axes-level function for histograms).
  warnings.warn(msg, FutureWarning)

df['session_number_'].plot.box(figsize=(5,10))

df.groupby('ab_cohort_name')['user_id'].count()

df.groupby("ab_cohort_name").publisher_revenue.describe()

	countfloat64	meanfloat64	stdfloat64	minfloat64	25%float64	50%float64	75%float64	maxfloat64
control	62952	0.03796353335226838	0.05837581486983601	0	0.0018903	0.017654200000000002	0.04994129974999997	1.265953403
gameTune	61238	0.03945714717650803	0.061136369158347414	0	0.0019642	0.0176653999999999	0.051940475	1.1712859999999985
xHigh	62103	0.04021031855364473	0.06327390071805815	0	0.00203	0.018502	0.052873179	2.635761600000001
xLow	62517	0.034448674933682005	0.05322391639056267	0	0.00178924	0.015178	0.0450861	1.1126336179999994
xxHigh	61185	0.0408248700623682	0.06622444836660722	0	0.0019004	0.01824	0.053219802	1.9031350999999983
xxLow	62839	0.033200609626235285	0.053487214901567395	0	0.0018681999999998998	0.0145183789999999	0.043252099499999946	2.891823927000002
6 rows × 8 columns

df.groupby("ab_cohort_name").session_number_.describe()

	countfloat64	meanfloat64	stdfloat64	minfloat64	25%float64	50%float64	75%float64	maxfloat64
control	62952	4.97809442114627	5.970849416190151	1	1	3	6	91
gameTune	61238	4.593422384793755	5.088682378649511	1	1	3	6	61
xHigh	62103	4.691915044361786	5.283734611684529	1	1	3	6	76
xLow	62517	4.769086808388119	6.126662988839299	0	1	3	6	137
xxHigh	61185	4.704976709977935	5.493924847322759	1	1	3	6	62
xxLow	62839	4.659319849138274	5.181069576723725	1	1	3	6	63
6 rows × 8 columns

# Let's create different datasets for each groups we want to test. data_control = df[df['ab_cohort_name'] == "control"] #only users that are in the control group. data_xHigh = df[df['ab_cohort_name'] == "xHigh"] #only users that are in the xHigh group. data_xxHigh = df[df['ab_cohort_name'] == "xxHigh"] #only users that are in the xxHigh group.

#We first compare the 'control' group's revenue with those generated from the 'xHigh' group. t_stat, p_val= stats.ttest_ind(data_control['publisher_revenue'], data_xHigh['publisher_revenue']) t_stat , p_val

#Then we compare the 'control' group's revenue with those generated from the 'xxHigh' group. t_stat, p_val= stats.ttest_ind(data_control['publisher_revenue'], data_xxHigh['publisher_revenue']) t_stat , p_val

#Finally we compare the 'xHigh' group's revenue with those generated from the 'xxHigh' group. t_stat, p_val= stats.ttest_ind(data_xHigh['publisher_revenue'], data_xxHigh['publisher_revenue']) t_stat , p_val