/
simulator.py
181 lines (121 loc) · 6.19 KB
/
simulator.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
import numpy as np
import itertools as it
import functools as ft
import pandas as pd
from scipy.stats import binom_test
from tqdm import tqdm
N_POSITIONS = [500, 350, 200, 150, 100, 75, 40, 10]
LEAVING_PROPORTION = 0.15
def generate_employee(level, founder=False, effect_size=None,
p=None, mu=50, sigma=10):
if effect_size is None:
effect_size = 0
if p is None:
p = 0.5
while True:
gender = 'Male' if np.random.rand() < p else 'Female'
shift = effect_size * sigma if gender == 'Female' else 0
score = np.random.normal(mu, sigma) + shift
yield {'level': level,
'founder': founder,
'gender': gender,
'score': score}
def initialize(n_positions, employees_generator):
employees_by_levels = (it.islice(employees_generator(level=level,
founder=True),
n_position_in_level)
for level, n_position_in_level
in enumerate(n_positions))
df_by_levels = [pd.DataFrame(employees) for employees in employees_by_levels]
employees_df = pd.concat(df_by_levels,
ignore_index=True)
return employees_df
def leave(employees_df, leaving_proportion):
employees_df = employees_df.sample(frac=1-leaving_proportion)
return employees_df
def promote_level(employees_df, level, n_positions):
upper_level = employees_df[employees_df['level'] == level]
lower_level = employees_df[employees_df['level'] == level-1]
n_to_promote = n_positions[level] - len(upper_level)
index_to_promote = lower_level.nlargest(n_to_promote, 'score').index
employees_df.loc[index_to_promote, 'level'] = level
return employees_df
def recruit(employees_df, n_positions, employees_generator):
first_level = employees_df[employees_df['level'] == 0]
n_to_recruit = n_positions[0] - len(first_level)
new_employees = it.islice(employees_generator(level=0,
founder=False),
n_to_recruit)
new_employees_df = pd.DataFrame(new_employees)
employees_df = employees_df.append(new_employees_df)
return employees_df
def simulate_timestep(employees_df, leaving_proportion, n_positions,
effect_size=None):
employees_df = leave(employees_df, leaving_proportion)
for level in range(len(n_positions)-1, 0, -1):
employees_df = promote_level(employees_df, level, n_positions)
employees_df = recruit(employees_df, n_positions, effect_size)
employees_df = employees_df.sort_values(by='level')
employees_df = employees_df.reset_index(drop=True)
return employees_df
def simulate_until_no_founder(n_positions, leaving_proportion,
effect_size=None, p=None):
employees_generator = ft.partial(generate_employee,
effect_size=effect_size,
p=p)
employees_df = initialize(n_positions, employees_generator)
with tqdm(total=float('inf')) as pbar:
while employees_df['founder'].any():
employees_df = simulate_timestep(employees_df,
leaving_proportion,
n_positions,
employees_generator)
pbar.update(1)
return employees_df
def calc_binom_pvalue(female_proportion, positions, p):
return binom_test(female_proportion * positions,
positions,
1-p)
def generate_one_simulation_results(n_positions,
leaving_proportion,
p=None,
effect_size=None):
employees_df = simulate_until_no_founder(n_positions,
leaving_proportion,
p=p,
effect_size=effect_size)
employees_by_level = employees_df.groupby('level')
simulation_result = pd.concat({'positions': employees_by_level.size(),
'mean': employees_by_level['score'].mean(),
'female_proportion': (employees_by_level['gender']
.apply(lambda g: (g == 'Female').mean())),
}, axis=1).sort_index(ascending=False)
simulation_result['pvalue'] = (simulation_result
.apply(lambda r:
calc_binom_pvalue(
r['female_proportion'],
r['positions'],
p),
axis=1)
)
simulation_result = simulation_result[['positions',
'mean',
'female_proportion',
'pvalue']]
return simulation_result
def run_simulation(effect_size=0, male_prop=0.5, random_seed=None):
if random_seed is not None:
np.random.seed(random_seed)
simulation_result = generate_one_simulation_results(N_POSITIONS,
LEAVING_PROPORTION,
effect_size=effect_size,
p=male_prop)
simulation_result = simulation_result.round(2)
simulation_result.index.name = simulation_result.index.name.title()
simulation_result.columns = [name.replace('_', ' ').title()
for name in simulation_result.columns]
gender_proportions = {'Female': simulation_result['Female Proportion'],
'Male': 1-simulation_result['Female Proportion']}
(pd.DataFrame(gender_proportions)[::-1]
.plot(kind='barh', stacked=True))
return simulation_result