softblock/run_test.py at master · ddimmery/softblock · GitHub

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
import pandas as pd
from tqdm import tqdm

import dgp
import design
from plan import Plan
import estimator as est
import evaluator as evalr
import numpy as np

def make_plan(designs):
    plan = Plan()

    for name, dgn, estr in designs:
        plan.add_design(name, dgn, estr)

    plan.add_evaluator('ATEError', evalr.ATEError)
    plan.add_evaluator('ITEBias', evalr.ITEBias)
    plan.add_evaluator('ITEMSE', evalr.ITEMSE)
    plan.add_evaluator('CovariateMSE', evalr.CovariateMSE)
    plan.add_evaluator('ATECovers', evalr.ATECovers)
    plan.add_evaluator('CISize', evalr.CISize)
    return plan

NUM_ITERS = 100

plan = make_plan([
    ('QuickBlock-B', design.QuickBlock, est.Blocking),
    ('QuickBlock-RF', design.QuickBlock, est.BlockingRF),
    ('OptBlock-B', design.OptBlock, est.Blocking),
    ('OptBlock-RF', design.OptBlock, est.BlockingRF),
    ('Randomization-RF', design.Bernoulli, est.OLSandRFT),
    ('Fixed Margins Randomization-RF', design.Complete, est.OLSandRFT),
    ('Rerandomization-RF', design.ReRandomization, est.OLSandRFT),
    #('Rerandomization-KNN', design.ReRandomization, est.OLSandKNNT),
    ('Matched Pair-B', design.MatchedPair, est.MatchedPairBlocking),
    ('SoftBlock-L', design.SoftBlock, est.LaplacianNorm),
    ('SoftBlock-RF', design.SoftBlock, est.OLSandRFT),
    ('GreedyNeighbors-RF', design.GreedyNeighbors, est.OLSandRFT),
    ('GreedyNeighbors-L', design.GreedyNeighbors, est.LaplacianNorm),
    #('KallusHeuristic-KNN', design.Heuristic, est.DMandKNNT),
    ('KallusHeuristic-RF', design.Heuristic, est.DMandRFT),
    #('KallusPSOD-KNN', design.PSOD, est.DMandKNNT),
    ('KallusPSOD-RF', design.PSOD, est.DMandRFT),
])

dfs = []
dgp_factory_class_list = [dgp.LinearFactory, dgp.QuickBlockFactory, dgp.SinusoidalFactory, dgp.TwoCirclesFactory]
sample_sizes = [10, 20, 32, 50, 64, 72, 86, 100]
for sample_size in sample_sizes[::-1]:
    print(f"Sample Size: {sample_size}")
    dgp_factory_list = [factory(N=sample_size) for factory in dgp_factory_class_list]
    for dgp_factory in dgp_factory_list:
        dgp_name = type(dgp_factory.create_dgp()).__name__
        print(f"DGP name: {dgp_name}")
        for it in tqdm(range(NUM_ITERS)):
            result = plan.execute(dgp_factory, seed=it * 1001)
            result['iteration'] = it
            result['sample_size'] = sample_size
            result['dgp'] = dgp_name
            filename = f"results/{dgp_name}_n{sample_size}_i{it}.csv.gz"
            result.to_csv(filename, index=False)
            dfs.append(result)


plan = make_plan([
    ('QuickBlock-B', design.QuickBlock, est.Blocking),
    ('QuickBlock-RF', design.QuickBlock, est.BlockingRF),
    #('OptBlock-B', design.OptBlock, est.Blocking),
    ('OptBlock-RF', design.OptBlock, est.BlockingRF),
    ('SoftBlock-L', design.SoftBlock, est.LaplacianNorm),
    ('SoftBlock-RF', design.SoftBlock, est.OLSandRFT),
    ('GreedyNeighbors-RF', design.GreedyNeighbors, est.OLSandRFT),
    ('GreedyNeighbors-L', design.GreedyNeighbors, est.LaplacianNorm),
    ('Randomization-RF', design.Bernoulli, est.OLSandRFT),
    ('Fixed Margins Randomization-RF', design.Bernoulli, est.OLSandRFT),
    ('Rerandomization-RF', design.ReRandomization, est.OLSandRFT),
    #('Rerandomization-KNN', design.ReRandomization, est.OLSandKNNT),
    #('KallusPSOD-KNN', design.PSOD, est.DMandKNNT),
    ('KallusPSOD-RF', design.PSOD, est.DMandRFT),
])

sample_sizes = [128, 150, 200, 250, 500, 1000, 2000, 3000]
for sample_size in sample_sizes[::-1]:
    print(f"Sample Size: {sample_size}")
    dgp_factory_list = [factory(N=sample_size) for factory in dgp_factory_class_list]
    for dgp_factory in dgp_factory_list:
        dgp_name = type(dgp_factory.create_dgp()).__name__
        print(f"DGP name: {dgp_name}")
        for it in tqdm(range(NUM_ITERS)):
            result = plan.execute(dgp_factory, seed=it * 1001)
            result['iteration'] = it
            result['sample_size'] = sample_size
            result['dgp'] = dgp_name
            filename = f"results/{dgp_name}_n{sample_size}_i{it}.csv.gz"
            result.to_csv(filename, index=False)
            dfs.append(result)


plan = make_plan([
    ('SoftBlock-L', design.SoftBlock, est.LaplacianNorm),
    ('SoftBlock-RF', design.SoftBlock, est.OLSandRFT),
    ('GreedyNeighbors-RF', design.GreedyNeighbors, est.OLSandRFT),
    ('GreedyNeighbors-L', design.GreedyNeighbors, est.LaplacianNorm),
    ('QuickBlock-B', design.QuickBlock, est.Blocking),
    ('QuickBlock-RF', design.QuickBlock, est.BlockingRF),
    ('Randomization', design.Bernoulli, est.OLSandRFT),
    ('Fixed Margins Randomization-RF', design.Complete, est.OLSandRFT),
    ('Rerandomization-RF', design.ReRandomization, est.OLSandRFT),
    # ('Rerandomization-KNN', design.ReRandomization, est.OLSandKNNT),
])

NUM_ITERS = 25

sample_sizes = [4000, 5000, 7500, 10000, 12500, 15000, 20000, 25000, 30000, 50000, 100000]
for sample_size in sample_sizes:
    print(f"Sample Size: {sample_size}")
    dgp_factory_list = [factory(N=sample_size) for factory in dgp_factory_class_list]
    for dgp_factory in dgp_factory_list:
        dgp_name = type(dgp_factory.create_dgp()).__name__
        print(f"DGP name: {dgp_name}")
        for it in tqdm(range(NUM_ITERS)):
            result = plan.execute(dgp_factory, seed=it * 1001)
            result['iteration'] = it
            result['sample_size'] = sample_size
            result['dgp'] = dgp_name
            filename = f"results/{dgp_name}_n{sample_size}_i{it}.csv.gz"
            result.to_csv(filename, index=False)
            dfs.append(result)


NUM_ITERS = 5

sample_sizes = [50000, 100000, 500000, 1000000]
for sample_size in sample_sizes:
    print(f"Sample Size: {sample_size}")
    dgp_factory_list = [factory(N=sample_size) for factory in dgp_factory_class_list]
    for dgp_factory in dgp_factory_list:
        dgp_name = type(dgp_factory.create_dgp()).__name__
        print(f"DGP name: {dgp_name}")
        for it in tqdm(range(NUM_ITERS)):
            result = plan.execute(dgp_factory, seed=it * 1001)
            result['iteration'] = it
            result['sample_size'] = sample_size
            result['dgp'] = dgp_name
            filename = f"results/{dgp_name}_n{sample_size}_i{it}.csv.gz"
            result.to_csv(filename, index=False)
            dfs.append(result)

results = pd.concat(dfs)

filename = f"results/all_results.csv.gz"

print(f"""
\n**********************************************************************
***\tSAVING TO `{filename}`\t\t   ***
**********************************************************************""")

results.to_csv(filename, index=False)