|
1
|
|
|
import math |
|
2
|
|
|
|
|
3
|
|
|
|
|
4
|
|
|
class Observation(object): |
|
5
|
|
|
x = None |
|
6
|
|
|
y = None |
|
7
|
|
|
group_id = None |
|
8
|
|
|
label = None |
|
9
|
|
|
|
|
10
|
|
|
def __init__(self, x=None, label=None, group_id=None, y=None): |
|
11
|
|
|
if x is not None: |
|
12
|
|
|
self.x = x |
|
13
|
|
|
if y is not None: |
|
14
|
|
|
self.y = y |
|
15
|
|
|
|
|
16
|
|
|
if label is not None: |
|
17
|
|
|
self.label = label |
|
18
|
|
|
|
|
19
|
|
|
if group_id is not None: |
|
20
|
|
|
self.group_id = group_id |
|
21
|
|
|
|
|
22
|
|
|
def is_categorical(self): |
|
23
|
|
|
return self.label is not None |
|
24
|
|
|
|
|
25
|
|
|
def is_numerical(self): |
|
26
|
|
|
return self.x is not None |
|
27
|
|
|
|
|
28
|
|
|
|
|
29
|
|
|
class Sample(object): |
|
30
|
|
|
observations = None |
|
31
|
|
|
|
|
32
|
|
|
def __init__(self): |
|
33
|
|
|
self.observations = [] |
|
34
|
|
|
|
|
35
|
|
|
def add(self, observation): |
|
36
|
|
|
self.observations.append(observation) |
|
37
|
|
|
|
|
38
|
|
|
def add_numeric(self, numeric, group_id=None): |
|
39
|
|
|
ob = Observation() |
|
40
|
|
|
ob.x = numeric |
|
41
|
|
|
ob.group_id = group_id |
|
42
|
|
|
self.add(ob) |
|
43
|
|
|
|
|
44
|
|
|
def add_category(self, category, group_id=None): |
|
45
|
|
|
ob = Observation() |
|
46
|
|
|
ob.label = category |
|
47
|
|
|
ob.group_id = group_id |
|
48
|
|
|
self.add(ob) |
|
49
|
|
|
|
|
50
|
|
|
def add_xy(self, x, y, group_id=None): |
|
51
|
|
|
ob = Observation() |
|
52
|
|
|
ob.x = x |
|
53
|
|
|
ob.y = y |
|
54
|
|
|
ob.group_id = group_id |
|
55
|
|
|
self.add(ob) |
|
56
|
|
|
|
|
57
|
|
|
def size(self): |
|
58
|
|
|
return len(self.observations) |
|
59
|
|
|
|
|
60
|
|
|
def get(self, index): |
|
61
|
|
|
return self.observations[index] |
|
62
|
|
|
|
|
63
|
|
|
def is_categorical(self): |
|
64
|
|
|
return self.observations[0].is_categorical() |
|
65
|
|
|
|
|
66
|
|
|
def is_numerical(self): |
|
67
|
|
|
return self.observations[0].is_numerical() |
|
68
|
|
|
|
|
69
|
|
|
def count_by_group_id(self, group_id): |
|
70
|
|
|
return sum(1 for x in self.observations if group_id is None or x.group_id == group_id) |
|
71
|
|
|
|
|
72
|
|
|
|
|
73
|
|
|
class SampleDistribution(object): |
|
74
|
|
|
sample = None |
|
75
|
|
|
group_id = None |
|
76
|
|
|
|
|
77
|
|
|
categorical_value = None |
|
78
|
|
|
is_categorical = False |
|
79
|
|
|
is_numerical = False |
|
80
|
|
|
|
|
81
|
|
|
sd = None |
|
82
|
|
|
sample_size = None |
|
83
|
|
|
mean = None |
|
84
|
|
|
variance = None |
|
85
|
|
|
sum_of_squares= None |
|
86
|
|
|
|
|
87
|
|
|
proportion = None |
|
88
|
|
|
|
|
89
|
|
|
def __init__(self, sample=None, group_id=None, categorical_value=None, mean=None, sd=None, sample_size=None, proportion=None): |
|
90
|
|
|
if group_id is not None: |
|
91
|
|
|
self.group_id = group_id |
|
92
|
|
|
if categorical_value is not None: |
|
93
|
|
|
self.categorical_value = categorical_value |
|
94
|
|
|
|
|
95
|
|
|
if mean is not None: |
|
96
|
|
|
self.mean = mean |
|
97
|
|
|
self.is_numerical = True |
|
98
|
|
|
|
|
99
|
|
|
if proportion is not None: |
|
100
|
|
|
self.proportion = proportion |
|
101
|
|
|
self.is_categorical = True |
|
102
|
|
|
|
|
103
|
|
|
if sample_size is not None: |
|
104
|
|
|
self.sample_size = sample_size |
|
105
|
|
|
|
|
106
|
|
|
if sd is not None: |
|
107
|
|
|
self.sd = sd |
|
108
|
|
|
|
|
109
|
|
|
if self.sd is not None and self.sample_size is not None: |
|
110
|
|
|
self.variance = self.sd * self.sd |
|
111
|
|
|
self.sum_of_squares = self.variance * (self.sample_size-1) |
|
112
|
|
|
|
|
113
|
|
|
if sample is not None: |
|
114
|
|
|
self.sample = sample |
|
115
|
|
|
if sample.is_numerical(): |
|
116
|
|
|
self.mean = SampleDistribution.calculate_mean(sample, group_id) |
|
117
|
|
|
self.sum_of_squares = SampleDistribution.calculate_sum_of_squares(sample, self.mean, group_id) |
|
118
|
|
|
self.sample_size = sample.count_by_group_id(group_id) |
|
119
|
|
|
self.variance = self.sum_of_squares / (self.sample_size - 1) |
|
120
|
|
|
self.sd = math.sqrt(self.variance) |
|
121
|
|
|
self.is_numerical = True |
|
122
|
|
|
elif sample.is_categorical() and categorical_value is not None: |
|
123
|
|
|
self.proportion = SampleDistribution.calculate_proportion(sample, categorical_value, group_id) |
|
124
|
|
|
self.sample_size = sample.count_by_group_id(group_id) |
|
125
|
|
|
self.mean = self.proportion * self.sample_size |
|
126
|
|
|
self.variance = self.proportion * (1.0 - self.proportion) * self.sample_size |
|
127
|
|
|
self.is_categorical = True |
|
128
|
|
|
|
|
129
|
|
|
@staticmethod |
|
130
|
|
|
def calculate_mean(sample, group_id): |
|
131
|
|
|
count = 0 |
|
132
|
|
|
the_sum = 0 |
|
133
|
|
|
for i in range(sample.size()): |
|
134
|
|
|
observation = sample.get(i) |
|
135
|
|
|
if group_id is not None and observation.group_id != group_id: |
|
136
|
|
|
continue |
|
137
|
|
|
the_sum += observation.x |
|
138
|
|
|
count += 1 |
|
139
|
|
|
return the_sum / count |
|
140
|
|
|
|
|
141
|
|
|
@staticmethod |
|
142
|
|
|
def calculate_sum_of_squares(sample, mean, group_id): |
|
143
|
|
|
the_sum = 0 |
|
144
|
|
|
for i in range(sample.size()): |
|
145
|
|
|
observation = sample.get(i) |
|
146
|
|
|
if group_id is not None and observation.group_id != group_id: |
|
147
|
|
|
continue |
|
148
|
|
|
the_sum += (observation.x - mean) * (observation.x - mean) |
|
149
|
|
|
return the_sum |
|
150
|
|
|
|
|
151
|
|
|
@staticmethod |
|
152
|
|
|
def calculate_proportion(sample, categorical_value, group_id): |
|
153
|
|
|
counter1 = 0 |
|
154
|
|
|
counter2 = 0 |
|
155
|
|
|
for i in range(sample.size()): |
|
156
|
|
|
observation = sample.get(i) |
|
157
|
|
|
if group_id is not None and observation.group_id != group_id: |
|
158
|
|
|
continue |
|
159
|
|
|
counter2 += 1 |
|
160
|
|
|
if observation.label == categorical_value: |
|
161
|
|
|
counter1 += 1 |
|
162
|
|
|
if counter2 == 0: |
|
163
|
|
|
return 0 |
|
164
|
|
|
return counter1 / counter2 |
|
165
|
|
|
|
|
166
|
|
|
|
chen0040 /
pysie
