Hide keyboard shortcuts

Hot-keys on this page

r m x p   toggle line displays

j k   next/prev highlighted chunk

0   (zero) top of page

1   (one) first highlighted chunk

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

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

# Copyright 2017-2020 Spotify AB 

# 

# Licensed under the Apache License, Version 2.0 (the "License"); 

# you may not use this file except in compliance with the License. 

# You may obtain a copy of the License at 

# 

# http://www.apache.org/licenses/LICENSE-2.0 

# 

# Unless required by applicable law or agreed to in writing, software 

# distributed under the License is distributed on an "AS IS" BASIS, 

# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 

# See the License for the specific language governing permissions and 

# limitations under the License. 

 

from abc import ABCMeta, abstractmethod 

from functools import wraps 

import types 

import warnings 

 

import chartify 

import numpy as np 

import pandas as pd 

 

from spotify_confidence.options import options 

from spotify_confidence.chartgrid import ChartGrid 

 

# warnings.simplefilter("once") 

 

INITIAL_RANDOMIZATION_SEED = np.random.get_state()[1][0] 

 

 

def axis_format_precision(max_value, min_value, absolute): 

extra_zeros = 2 if absolute else 0 

precision = -int(np.log10(abs(max_value - min_value))) + extra_zeros 

zeros = ''.join(['0'] * precision) 

return "0.{}{}".format(zeros, '' if absolute else '%') 

 

 

def add_color_column(df, cols): 

for i, column in enumerate(cols): 

if i == 0: 

df['color'] = df[column] 

else: 

df['color'] = df['color'] + ' ' + df[column] 

return df 

 

 

def randomization_warning_decorator(f): 

"""Set numpy randomization seed and warn users if not fixed. 

 

Note to developers: 

Do not compare random variables that have been 

sampled from the same seed. It will lead to incorrect results. 

To avoid this situation it's best to apply this decorator to 

public methods that involve randomization. 

""" 

 

@wraps(f) 

def wrapper(*args, **kwargs): 

 

option_seed = options.get_option('randomization_seed') 

np_seed = INITIAL_RANDOMIZATION_SEED 

if option_seed != np_seed and option_seed is None: 

randomization_warning_message = """ 

Your analysis will not be reproducible! 

Using a method that involves randomization without setting a seed. 

Please run the following and add it to the top of your script or 

notebook after you import confidence: 

 

confidence.options.set_option('randomization_seed', {}) 

 

""".format(INITIAL_RANDOMIZATION_SEED) 

warnings.warn(randomization_warning_message) 

option_seed = np_seed 

np.random.seed(option_seed) 

return f(*args, **kwargs) 

 

return wrapper 

 

 

class BaseTest(object, metaclass=ABCMeta): 

"""Base test class that provides abstract methods 

to ensure consistency across test classes.""" 

 

def __init__(self, 

data_frame, 

categorical_group_columns, 

ordinal_group_column, 

numerator_column, 

denominator_column, 

interval_size): 

 

self._data_frame = data_frame 

self._numerator_column = numerator_column 

self._denominator_column = denominator_column 

self._interval_size = interval_size 

 

categorical_string_or_none = (isinstance(categorical_group_columns, 

str) 

or categorical_group_columns is None) 

self._categorical_group_columns = [ 

categorical_group_columns 

] if categorical_string_or_none else categorical_group_columns 

self._ordinal_group_column = ordinal_group_column 

 

self._all_group_columns = self._categorical_group_columns + [ 

self._ordinal_group_column 

] 

self._all_group_columns = [ 

column for column in self._all_group_columns if column is not None 

] 

self._validate_data() 

 

def _validate_data(self): 

"""Integrity check input dataframe. 

""" 

if not self._all_group_columns: 

raise ValueError("""At least one of `categorical_group_columns` 

or `ordinal_group_column` must be specified.""" 

) 

 

# Ensure there's at most 1 observation per grouping. 

max_one_row_per_grouping = all( 

self._data_frame.groupby(self._all_group_columns).size() <= 1) 

if not max_one_row_per_grouping: 

raise ValueError( 

"""Each grouping should have at most 1 observation.""") 

 

if self._ordinal_group_column: 

ordinal_column_type = self._data_frame[ 

self._ordinal_group_column].dtype.type 

if not np.issubdtype(ordinal_column_type, 

np.number) and not issubclass( 

ordinal_column_type, np.datetime64): 

raise TypeError("""`ordinal_group_column` is type `{}`. 

Must be number or datetime type.""".format(ordinal_column_type)) 

 

@classmethod 

def as_cumulative( 

cls, 

data_frame, 

numerator_column, 

denominator_column, 

ordinal_group_column, 

categorical_group_columns=None): 

""" 

Instantiate the class with a cumulative representation of the dataframe. 

Sorts by the ordinal variable and calculates the cumulative sum 

May be used for to visualize the difference between groups as a 

time series. 

 

Args: 

data_frame (pd.DataFrame): DataFrame 

numerator_column (str): Column name for numerator column. 

denominator_column (str): Column name for denominator column. 

ordinal_group_column (str): Column name for ordinal grouping 

(e.g. numeric or date values). 

categorical_group_columns (str or list), 

Optional: Column names for categorical groupings. 

 

""" 

 

sorted_df = data_frame.sort_values(by=ordinal_group_column) 

cumsum_cols = [numerator_column, denominator_column] 

if categorical_group_columns: 

sorted_df[cumsum_cols] = ( 

sorted_df.groupby(by=categorical_group_columns)[cumsum_cols] 

.cumsum()) 

else: 

sorted_df[cumsum_cols] = sorted_df[cumsum_cols].cumsum() 

 

return cls(sorted_df, numerator_column, denominator_column, 

categorical_group_columns, ordinal_group_column) 

 

def summary(self): 

"""Return Pandas DataFrame with summary statistics. 

""" 

return self._summary(self._data_frame, self._interval) 

 

def _summary(self, data_frame, ci_function): 

"""Return the input dataframe with added columns: 

- Lower & upper bounds of 

Bayesian: credible interval 

Frequentist: confidence interval 

- Additional summary stats 

(e.g. probability in the case of Binomial data) 

""" 

summary_df = data_frame[ 

self._all_group_columns + 

[self._numerator_column, self._denominator_column]].copy() 

 

summary_df['point_estimate'] = \ 

(summary_df[self._numerator_column] * 1.0 / 

summary_df[self._denominator_column]) 

summary_df[['ci_lower', 'ci_upper']] = \ 

(data_frame.apply(ci_function, axis=1, result_type='expand')) 

 

return summary_df 

 

def summary_plot(self, groupby=None): 

"""Plot for each group in the data_frame: 

 

if ordinal level exists: 

Frequentist: line graph with area to represent confidence interval 

Bayesian: line graph with area to represent credible interval 

if categorical levels: 

Bayesian: KDE plot of posterior distributions by group 

Frequentist: Interval plots of confidence intervals by group 

 

Args: 

groupby (str): Name of column. 

If specified, will plot a separate chart for each level of the 

grouping. 

 

Returns: 

ChartGrid object. 

""" 

chart_grid = self._iterate_groupby_to_chartgrid(self._summary_plot, 

groupby=groupby) 

return chart_grid 

 

def _summary_plot(self, level_name, level_df, remaining_groups, groupby): 

 

if (self._ordinal_group_column is not None 

and self._ordinal_group_column in remaining_groups): 

 

ch = self._ordinal_summary_plot(level_name, level_df, 

remaining_groups, groupby) 

else: 

ch = self._categorical_summary_plot(level_name, level_df, 

remaining_groups, groupby) 

return ch 

 

def _ordinal_summary_plot(self, level_name, level_df, 

remaining_groups, groupby): 

remaining_groups = self._remaining_categorical_groups(remaining_groups) 

df = self._summary(level_df, self._interval) 

title = "Estimate of {} / {}".format(self._numerator_column, 

self._denominator_column) 

y_axis_label = "{} / {}".format(self._numerator_column, 

self._denominator_column) 

return self._ordinal_plot('point_estimate', df, groupby, level_name, 

remaining_groups, absolute=True, 

title=title, y_axis_label=y_axis_label) 

 

def _ordinal_plot(self, center_name, df, groupby, level_name, 

remaining_groups, absolute, title, y_axis_label): 

df = add_color_column(df, remaining_groups) 

colors = 'color' if remaining_groups else None 

ch = chartify.Chart(x_axis_type=self._ordinal_type()) 

ch.plot.line( 

data_frame=df.sort_values(self._ordinal_group_column), 

x_column=self._ordinal_group_column, 

y_column=center_name, 

color_column=colors) 

ch.style.color_palette.reset_palette_order() 

ch.plot.area( 

data_frame=df.sort_values(self._ordinal_group_column), 

x_column=self._ordinal_group_column, 

y_column='ci_lower', 

second_y_column='ci_upper', 

color_column=colors) 

ch.axes.set_yaxis_label(y_axis_label) 

ch.axes.set_xaxis_label(self._ordinal_group_column) 

ch.set_source_label("") 

axis_format = axis_format_precision(df['ci_lower'].min(), 

df['ci_upper'].max(), absolute) 

ch.axes.set_yaxis_tick_format(axis_format) 

subtitle = "" if not groupby else "{}: {}".format(groupby, level_name) 

ch.set_subtitle(subtitle) 

ch.set_title(title) 

if colors: 

ch.set_legend_location('outside_bottom') 

return ch 

 

def _remaining_categorical_groups(self, remaining_groups): 

remaining_groups_list = [remaining_groups] \ 

if isinstance(remaining_groups, str) else remaining_groups 

 

remaining_categorical_groups = [ 

group_name for group_name in remaining_groups_list 

if group_name != self._ordinal_group_column 

] 

return remaining_categorical_groups 

 

def _ordinal_type(self): 

ordinal_column_type = \ 

self._data_frame[self._ordinal_group_column].dtype.type 

axis_type = 'datetime' if issubclass(ordinal_column_type, 

np.datetime64) else 'linear' 

return axis_type 

 

@abstractmethod 

def _categorical_summary_plot(self, 

level_name, 

level_df, 

remaining_groups, 

groupby): 

pass 

 

@abstractmethod 

def difference(self, level_1, level_2, absolute=True, groupby=None): 

"""Return dataframe containing the difference in means between 

group 1 and 2 and the appropriate test statistics. 

Frequentist: 

- Calculate one of the following tests depending of the 

response variable type. 

- Binomial: Chisq / fisher exact test 

- Gaussian: t-test / z-test 

Return the p-value. 

Bayesian: 

- Calcuate the posterior distribution of the difference in means. 

Return the 

- probability that group 2 > group 1. 

- Expected loss 

- Expected change 

- Expected gain 

- 95% CI interval 

""" 

pass 

 

def difference_plot(self, level_1, level_2, absolute=True, groupby=None): 

"""Plot representing the difference between group 1 and 2. 

- Difference in means or proportions, depending 

on the response variable type. 

 

Frequentist: 

- Plot interval plot with confidence interval of the 

difference between groups 

 

Bayesian: 

- Plot KDE representing the posterior distribution of the difference. 

- Probability that group2 > group1 

- Mean difference 

- 95% interval. 

 

Args: 

level_1 (str, tuple of str): Name of first level. 

level_2 (str, tuple of str): Name of second level. 

absolute (bool): If True then return the absolute 

difference (level2 - level1) 

otherwise return the relative difference (level2 / level1 - 1) 

groupby (str): Name of column, or list of columns. 

If specified, will return an interval for each level 

of the grouped dimension, or a confidence band if the 

grouped dimension is ordinal 

 

Returns: 

GroupedChart object. 

""" 

 

use_ordinal_axis = self._use_ordinal_axis(groupby) 

 

if use_ordinal_axis: 

ch = self._ordinal_difference_plot(level_1, level_2, 

absolute, groupby) 

chart_grid = ChartGrid() 

chart_grid.charts.append(ch) 

else: 

chart_grid = self._categorical_difference_plot(level_1, level_2, 

absolute, groupby) 

 

return chart_grid 

 

def _use_ordinal_axis(self, groupby): 

is_ordinal_difference_plot = ( 

groupby is not None and 

self._ordinal_group_column is not None and 

self._ordinal_group_column in groupby) 

return is_ordinal_difference_plot 

 

def _ordinal_difference_plot(self, level_1, level_2, absolute, groupby): 

difference_df = self.difference(level_1, level_2, 

absolute, groupby) 

remaining_groups = self._remaining_categorical_groups(groupby) 

title = "Change from {} to {}".format(level_1, level_2) 

y_axis_label = self.get_difference_plot_label(absolute) 

ch = self._ordinal_plot('difference', difference_df, groupby=None, 

level_name="", 

remaining_groups=remaining_groups, 

absolute=absolute, 

title=title, y_axis_label=y_axis_label) 

ch.callout.line(0) 

 

return ch 

 

def get_difference_plot_label(self, absolute): 

change_type = 'Absolute' if absolute else "Relative" 

return change_type + " change in {} / {}".format( 

self._numerator_column, self._denominator_column) 

 

@abstractmethod 

def _categorical_difference_plot(self, level_1, level_2, absolute, groupby): 

pass 

 

@abstractmethod 

def multiple_difference(self, 

level, 

absolute=True, 

groupby=None, 

level_as_reference=False): 

"""The pairwise probability that the specific group 

is greater than all other groups. 

""" 

pass 

 

def multiple_difference_plot(self, level, absolute=True, groupby=None, 

level_as_reference=False): 

"""Compare level to all other groups or, if level_as_reference = True, 

all other groups to level. 

 

Args: 

level (str, tuple of str): Name of level. 

absolute (bool): If True then return the absolute 

difference (level2 - level1) 

otherwise return the relative difference (level2 / level1 - 1) 

groupby (str): Name of column, or list of columns. 

If specified, will return an interval for each level 

of the grouped dimension, or a confidence band if the 

grouped dimension is ordinal 

level_as_reference: If false (default), compare level to all other 

groups. If true, compare all other groups to level. 

""" 

use_ordinal_axis = self._use_ordinal_axis(groupby) 

 

if use_ordinal_axis: 

ch = self._ordinal_multiple_difference_plot( 

level, 

absolute, 

groupby, 

level_as_reference) 

chart_grid = ChartGrid() 

chart_grid.charts.append(ch) 

else: 

chart_grid = self._categorical_multiple_difference_plot( 

level, 

absolute, 

groupby, 

level_as_reference) 

 

return chart_grid 

 

def _ordinal_multiple_difference_plot(self, level, absolute, groupby, 

level_as_reference): 

difference_df = self.multiple_difference(level, absolute, groupby, 

level_as_reference) 

remaining_groups = self._remaining_categorical_groups(groupby) 

groupby_columns = self._add_level_column(remaining_groups, 

level_as_reference) 

title = "Comparison to {}".format(level) 

y_axis_label = self.get_difference_plot_label(absolute) 

ch = self._ordinal_plot('difference', difference_df, groupby=None, 

level_name="", remaining_groups=groupby_columns, 

absolute=absolute, 

title=title, y_axis_label=y_axis_label) 

ch.callout.line(0) 

return ch 

 

def _add_level_column(self, groupby, level_as_reference): 

level_column = 'level_2' if level_as_reference else 'level_1' 

if groupby is None: 

groupby_columns = level_column 

else: 

if isinstance(groupby, str): 

groupby_columns = [groupby, level_column] 

else: 

groupby_columns = groupby + [level_column] 

return groupby_columns 

 

@abstractmethod 

def _categorical_multiple_difference_plot(self, level, absolute, 

groupby, level_as_reference): 

pass 

 

@staticmethod 

def _validate_levels(level_df, remaining_groups, level): 

try: 

level_df.groupby(remaining_groups).get_group(level) 

except (KeyError, ValueError): 

raise ValueError(""" 

Invalid level: '{}' 

Must supply a level within the ungrouped dimensions: {} 

Valid levels: 

{} 

""".format( 

level, remaining_groups, 

list(level_df.groupby(remaining_groups).groups.keys()))) 

 

def _groupby_iterator(self, input_function, groupby, **kwargs): 

groupby = [] if groupby is None else groupby 

# Will group over the whole dataframe if groupby is None 

level_groups = groupby if groupby else np.ones(len(self._data_frame)) 

 

remaining_groups = [ 

group for group in self._all_group_columns 

if group not in groupby and group is not None 

] 

 

for level_name, level_df in self._data_frame.groupby(level_groups): 

yield input_function(level_name, level_df, remaining_groups, 

groupby, **kwargs) 

 

def _iterate_groupby_to_chartgrid(self, input_function, groupby, **kwargs): 

"""Iterate through groups in the test and apply the input function. 

 

Returns ChartGrid""" 

chart_grid = ChartGrid() 

 

chart_grid.charts = list( 

self._groupby_iterator(input_function, groupby, **kwargs)) 

 

return chart_grid 

 

def _iterate_groupby_to_dataframe(self, input_function, groupby, **kwargs): 

"""Iterate through groups in the test and apply the input function. 

 

Returns pd.DataFrame""" 

groupby_iterator = self._groupby_iterator(input_function, groupby, 

**kwargs) 

 

# Flatten any nested generators. 

groupby_iterator = list(groupby_iterator) 

if isinstance(groupby_iterator[0], types.GeneratorType): 

groupby_iterator = [ 

group for generator in groupby_iterator for group in generator 

] 

 

results_data_frame = pd.concat(groupby_iterator, axis=0) 

 

results_data_frame = results_data_frame.reset_index(drop=True) 

 

return results_data_frame 

 

def _all_groups(self): 

"""Return a list of all group keys. 

 

Returns: list""" 

groups = list(self._data_frame.groupby(self._all_group_columns) 

.groups.keys()) 

return groups 

 

def _add_group_by_columns(self, difference_df, groupby, level_name): 

if groupby: 

groupby = groupby[0] if len(groupby) == 1 else groupby 

if isinstance(groupby, str): 

difference_df.insert(0, column=groupby, value=level_name) 

else: 

for col, val in zip(groupby, level_name): 

difference_df.insert(0, column=col, value=val) 

 

# class BinomialResponse(BaseTest, metaclass=ABCMeta): 

# """Binomial Response Variable. 

# """ 

 

# class GaussianResponse(BaseTest, metaclass=ABCMeta): 

# """Base class for tests of normal response variables 

 

# E.g. Revenue per user 

# """ 

 

# pass 

 

 

# class PoissonResponse(BaseTest, metaclass=ABCMeta): 

# """Base class for tests of poisson response variables. 

 

# E.g. # of days active per user per month 

# """ 

# pass 

 

 

# class MultinomialResponse(BaseTest, metaclass=ABCMeta): 

# """Base class for tests of multinomial response variables. 

 

# E.g. single choice answer survey 

# self. 

# """ 

 

# def __init__(self, data_frame, categorical_group_columns, 

# ordinal_group_column, category_column, value_column): 

# self._category_column = category_column 

# self._value_column = value_column 

# super().__init__(data_frame, categorical_group_columns, 

# ordinal_group_column) 

 

 

# class CategoricalResponse(BaseTest, metaclass=ABCMeta): 

# """Base class for tests of categorical response variables. 

 

# E.g. multiple choice answer survey 

# """ 

 

# def __init__(self, data_frame, categorical_group_columns, 

# ordinal_group_column, category_column, value_column): 

# self._category_column = category_column 

# self._value_column = value_column 

# super().__init__(data_frame, categorical_group_columns, 

# ordinal_group_column) 

 

# pass