chen0040 /
pysie
| @@ 37-70 (lines=34) @@ | ||
| 34 | self.p_value_two_tail < significance_level) |
|
| 35 | ||
| 36 | ||
| 37 | class ProportionDiffTesting(object): |
|
| 38 | sampling_distribution = None |
|
| 39 | p_value_one_tail = None |
|
| 40 | p_value_two_tail = None |
|
| 41 | mean_null = None |
|
| 42 | test_statistic = None |
|
| 43 | significance_level = None |
|
| 44 | reject_proportion_same = None |
|
| 45 | ||
| 46 | def __init__(self, sampling_distribution, significance_level=None): |
|
| 47 | self.sampling_distribution = sampling_distribution |
|
| 48 | p_null = (sampling_distribution.grp1_point_estimate + sampling_distribution.grp2_point_estimate) / 2 |
|
| 49 | self.p_null = p_null |
|
| 50 | if significance_level is not None: |
|
| 51 | self.significance_level = significance_level |
|
| 52 | ||
| 53 | if self.sampling_distribution.distribution_family == DistributionFamily.normal: |
|
| 54 | standard_error_null = math.sqrt(p_null * (1 - p_null) / sampling_distribution.sample_size) |
|
| 55 | Z = sampling_distribution.point_estimate / standard_error_null |
|
| 56 | self.test_statistic = Z |
|
| 57 | pf = norm.cdf(Z) |
|
| 58 | self.p_value_one_tail = 1 - pf |
|
| 59 | self.p_value_two_tail = self.p_value_one_tail * 2 |
|
| 60 | else: |
|
| 61 | standard_error_null = math.sqrt(p_null * (1 - p_null) / sampling_distribution.sample_size) |
|
| 62 | td_df = sampling_distribution.point_estimate / standard_error_null |
|
| 63 | self.test_statistic = td_df |
|
| 64 | pf = t.cdf(td_df, sampling_distribution.df) |
|
| 65 | self.p_value_one_tail = 1 - pf |
|
| 66 | self.p_value_two_tail = self.p_value_one_tail * 2 |
|
| 67 | ||
| 68 | if significance_level is not None: |
|
| 69 | self.reject_proportion_same = (self.p_value_one_tail < significance_level, |
|
| 70 | self.p_value_two_tail < significance_level) |
|
| @@ 42-74 (lines=33) @@ | ||
| 39 | self.p_value_two_tail < significance_level) |
|
| 40 | ||
| 41 | ||
| 42 | class ProportionTesting(object): |
|
| 43 | sampling_distribution = None |
|
| 44 | p_value_one_tail = None |
|
| 45 | p_value_two_tail = None |
|
| 46 | mean_null = None |
|
| 47 | test_statistic = None |
|
| 48 | significance_level = None |
|
| 49 | reject_mean_null = None |
|
| 50 | ||
| 51 | def __init__(self, sampling_distribution, p_null, significance_level=None): |
|
| 52 | self.sampling_distribution = sampling_distribution |
|
| 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 - p_null) / standard_error_null |
|
| 60 | self.test_statistic = Z |
|
| 61 | pf = norm.cdf(Z) |
|
| 62 | self.p_value_one_tail = 1 - pf |
|
| 63 | self.p_value_two_tail = self.p_value_one_tail * 2 |
|
| 64 | else: |
|
| 65 | standard_error_null = math.sqrt(p_null * (1 - p_null) / sampling_distribution.sample_size) |
|
| 66 | td_df = (sampling_distribution.point_estimate - p_null) / standard_error_null |
|
| 67 | self.test_statistic = td_df |
|
| 68 | pf = t.cdf(td_df, sampling_distribution.df) |
|
| 69 | self.p_value_one_tail = 1 - pf |
|
| 70 | self.p_value_two_tail = self.p_value_one_tail * 2 |
|
| 71 | ||
| 72 | if significance_level is not None: |
|
| 73 | self.reject_mean_null = (self.p_value_one_tail < significance_level, |
|
| 74 | self.p_value_two_tail < significance_level) |
|
| 75 | ||
| 76 | ||
