|
1
|
|
|
from pysie.stats.distributions import DistributionFamily |
|
2
|
|
|
from scipy.stats import norm, t |
|
3
|
|
|
import math |
|
4
|
|
|
|
|
5
|
|
|
|
|
6
|
|
|
class MeanDiffTesting(object): |
|
7
|
|
|
sampling_distribution = None |
|
8
|
|
|
p_value_one_tail = None |
|
9
|
|
|
p_value_two_tail = None |
|
10
|
|
|
test_statistic = None |
|
11
|
|
|
significance_level = None |
|
12
|
|
|
reject_mean_same = None |
|
13
|
|
|
|
|
14
|
|
View Code Duplication |
def __init__(self, sampling_distribution, significance_level=None): |
|
|
|
|
|
|
15
|
|
|
self.sampling_distribution = sampling_distribution |
|
16
|
|
|
if significance_level is not None: |
|
17
|
|
|
self.significance_level = significance_level |
|
18
|
|
|
|
|
19
|
|
|
if self.sampling_distribution.distribution_family == DistributionFamily.normal: |
|
20
|
|
|
Z = sampling_distribution.point_estimate / sampling_distribution.standard_error |
|
21
|
|
|
self.test_statistic = Z |
|
22
|
|
|
pf = norm.cdf(Z) |
|
23
|
|
|
if Z < 0: |
|
24
|
|
|
pf = 1 - pf |
|
25
|
|
|
self.p_value_one_tail = 1 - pf |
|
26
|
|
|
self.p_value_two_tail = self.p_value_one_tail * 2 |
|
27
|
|
|
else: |
|
28
|
|
|
td_df = sampling_distribution.point_estimate / sampling_distribution.standard_error |
|
29
|
|
|
self.test_statistic = td_df |
|
30
|
|
|
pf = t.cdf(td_df, sampling_distribution.df) |
|
31
|
|
|
if td_df < 0: |
|
32
|
|
|
pf = 1 - pf |
|
33
|
|
|
self.p_value_one_tail = 1 - pf |
|
34
|
|
|
self.p_value_two_tail = self.p_value_one_tail * 2 |
|
35
|
|
|
|
|
36
|
|
|
if significance_level is not None: |
|
37
|
|
|
self.reject_mean_same = (self.p_value_one_tail < significance_level, |
|
38
|
|
|
self.p_value_two_tail < significance_level) |
|
39
|
|
|
|
|
40
|
|
|
|
|
41
|
|
|
class ProportionDiffTesting(object): |
|
42
|
|
|
sampling_distribution = None |
|
43
|
|
|
p_value_one_tail = None |
|
44
|
|
|
p_value_two_tail = None |
|
45
|
|
|
mean_null = None |
|
46
|
|
|
test_statistic = None |
|
47
|
|
|
significance_level = None |
|
48
|
|
|
reject_proportion_same = None |
|
49
|
|
|
|
|
50
|
|
View Code Duplication |
def __init__(self, sampling_distribution, significance_level=None): |
|
|
|
|
|
|
51
|
|
|
self.sampling_distribution = sampling_distribution |
|
52
|
|
|
p_null = (sampling_distribution.grp1_point_estimate + sampling_distribution.grp2_point_estimate) / 2 |
|
53
|
|
|
self.p_null = p_null |
|
54
|
|
|
if significance_level is not None: |
|
55
|
|
|
self.significance_level = significance_level |
|
56
|
|
|
|
|
57
|
|
|
if self.sampling_distribution.distribution_family == DistributionFamily.normal: |
|
58
|
|
|
standard_error_null = math.sqrt(p_null * (1 - p_null) / sampling_distribution.sample_size) |
|
59
|
|
|
Z = sampling_distribution.point_estimate / standard_error_null |
|
60
|
|
|
self.test_statistic = Z |
|
61
|
|
|
pf = norm.cdf(Z) |
|
62
|
|
|
if Z < 0: |
|
63
|
|
|
pf = 1 - pf |
|
64
|
|
|
self.p_value_one_tail = 1 - pf |
|
65
|
|
|
self.p_value_two_tail = self.p_value_one_tail * 2 |
|
66
|
|
|
else: |
|
67
|
|
|
standard_error_null = math.sqrt(p_null * (1 - p_null) / sampling_distribution.sample_size) |
|
68
|
|
|
td_df = sampling_distribution.point_estimate / standard_error_null |
|
69
|
|
|
self.test_statistic = td_df |
|
70
|
|
|
pf = t.cdf(td_df, sampling_distribution.df) |
|
71
|
|
|
if td_df < 0: |
|
72
|
|
|
pf = 1 - pf |
|
73
|
|
|
self.p_value_one_tail = 1 - pf |
|
74
|
|
|
self.p_value_two_tail = self.p_value_one_tail * 2 |
|
75
|
|
|
|
|
76
|
|
|
if significance_level is not None: |
|
77
|
|
|
self.reject_proportion_same = (self.p_value_one_tail < significance_level, |
|
78
|
|
|
self.p_value_two_tail < significance_level) |
chen0040 /
pysie
Loading history...