-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathIMMR.py
723 lines (600 loc) · 30.4 KB
/
IMMR.py
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
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
from sklearn.preprocessing import PolynomialFeatures
import statsmodels.stats.stattools as stattools
import statsmodels.stats.api as sms
import statsmodels.api as sm
import dash_bootstrap_components as dbc
from dash import dcc, html, dash_table
import dash
import plotly.graph_objects as go
from scipy import stats
import pandas as pd
import numpy as np
import random
import json
import app_fxns
def generate_outputs_iterative_multi_model_regression():
return html.Div(
id="Figure1",
children=[
dcc.Loading(
id="loading-fig1",
type="default",
fullscreen=False,
children=html.Div(id="figure1",
children=[dcc.Graph(id='figure_plot1'),
],
),
),
html.P("", id='fig1txt'),
],
style={'width': '100%',
'display': 'inline-block',
'background-color': '#f0f0f0',
'padding': '1%',
},
)
def control_card_iterative_multi_model_regression():
return html.Div(
id="control-card1",
children=[
html.H5("Explore relationships between multiple features at once",
style={'display': 'inline-block',
'width': '41.5%',
},
),
html.I(className="fas fa-question-circle fa-lg", id="target_select_vars",
style={'display': 'inline-block',
'width': '3%',
'color':'#99ccff',
},
),
dbc.Tooltip("As a default, these analyses are based on ordinary least squares " +
"regression (OLS). These analyses exclude categorical features, any" +
"features suspected of being dates or times, and any numeric features" +
"having less than 4 unique values.",
target="target_select_vars",
style = {'font-size': 12,
},
),
html.Hr(),
html.B("Choose one or more x-variables.",
style={'display': 'inline-block',
'vertical-align': 'top',
},
),
dcc.Dropdown(
id='xvar',
options=[{"label": i, "value": i} for i in []],
multi=True,
value=None,
style={'width': '100%',
},
),
html.Br(),
html.B("Choose one or more y-variables.",
style={'display': 'inline-block',
'vertical-align': 'top',
},
),
dcc.Dropdown(
id='yvar',
options=[{"label": i, "value": i} for i in []],
multi=True,
value=None,
style={'width': '100%',
},
),
html.Hr(),
html.Br(),
dbc.Button('Run regressions',
id='btn1',
n_clicks=0,
style={'width': '20%',
'font-size': 12,
"background-color": "#2a8cff",
'display': 'inline-block',
'margin-right': '2%',
},
),
dbc.Button("View results table",
id="open-iterative_multimodel_ols_table1",
style={
'background-color': '#2a8cff',
'width': '16%',
'font-size': 12,
'display': 'inline-block',
'margin-right': '2%',
},
),
dbc.Modal(
[dbc.ModalBody([html.Div(id='table_plot1'),
html.Br(),
html.P("", id='table1txt'),
],
),
dbc.ModalFooter(
dbc.Button("Click to Close",
id="close-iterative_multimodel_ols_table1",
className="ml-auto",
style={
"background-color": "#2a8cff",
'width': '30%',
'font-size': 14,
},
),
style={
"background-color": "#A0A0A0",
"display": "flex",
"justify-content": "center",
"align-items": "center",
},
),
],
id="modal-iterative_multimodel_ols_table1",
is_open=False,
centered=True,
autoFocus=True,
#size="xl",
fullscreen=True,
keyboard=True,
fade=True,
backdrop=True,
),
dbc.Button('Smart scale',
id='btn_ss',
n_clicks=0,
style={'width': '15%',
'font-size': 12,
"background-color": "#2a8cff",
'display': 'inline-block',
'margin-right': '2%',
},
),
html.I(className="fas fa-question-circle fa-lg",
id="ss1",
style={'display': 'inline-block',
'width': '3%',
'color':'#99ccff',
},
),
dbc.Tooltip("Skewed data can weaken analyses and visualizations. Click on 'Smart" +
"Scale' and the app will automatically detect and apply the best " +
"scaling for each skewed variable. Smart scaling will not necessarily" +
"improve the r-square. To remove the rescaling just click " +
"'SMART SCALE' again.",
target="ss1",
style = {'font-size': 12,
'display': 'inline-block',
},
),
dbc.Button('Run robust',
id='btn_robust',
n_clicks=0,
style={'width': '15%',
'font-size': 12,
"background-color": "#2a8cff",
'display': 'inline-block',
'margin-right': '2%',
},
),
html.I(className="fas fa-question-circle fa-lg",
id="robust1",
style={'display': 'inline-block',
'width': '3%',
'color':'#99ccff',
},
),
dbc.Tooltip("Outliers can weaken OLS regression. However, clicking on 'RUN ROBUST'" +
" will run a 'Robust Linear Model' via statsmodels. The r-square " +
"(for observed vs predicted) of a robust regression will likely be " +
"lower than the r-square of OLS regression. This is because the " +
"robust model is not chasing outliers. However, for non-outliers " +
"(the main trend), the robust model will be more accurate, stable, " +
"valid, and useful for predictions than an OLS model. To run regular " +
"OLS regression, simply click 'RUN ROBUST' again.",
target="robust1",
style = {'font-size': 12,
'display': 'inline-block',
},
),
html.P("",
id='rt0',
),
],
style={'margin-left': '1%',
'width': '98.5%',
},
)
def get_updated_results(n_clicks, smartscale, robust, df, cat_vars, xvars, yvars):
ctx1 = dash.callback_context
jd1 = json.dumps({'triggered': ctx1.triggered,})
jd1 = jd1[:50]
if 'rt4.children' in jd1:
return {}, {}, "", "", ""
if df is None or xvars is None or yvars is None or len(xvars) == 0 or len(yvars) == 0:
return {}, {}, "", "", ""
elif len(xvars) == 1 and len(yvars) == 1 and xvars[0] == yvars[0]:
return [{}, {}, "Error: Your predictor variable and response variable cannot be the same.",
"", ""]
else:
df = pd.DataFrame(df)
if df.empty:
return {}, {}, "", "", ""
df.drop(labels=cat_vars, axis=1, inplace=True)
vars_ = xvars + yvars
vars_ = list(set(vars_))
df = df.filter(items=vars_, axis=1)
if df.shape[0] == 0:
return [{}, {},
"Error: There are no rows in the data because of the variables you chose.",
"", ""]
else:
if smartscale % 2 != 0:
df, xvars, yvars = app_fxns.smart_scale(df, xvars, yvars)
models = []
coefs = []
eqns = []
r2s = []
adj_r2s = []
obs_pred_r2s = []
intercepts = []
pvals = []
bics = []
aics = []
ns = []
Yvars = []
Yvars_short = []
Xvars = []
Xvars_short = []
llf_ls = []
Xs = []
Ys = []
PredY = []
durbin_watson = []
breusch_pagan = []
jarque_bera = []
harvey_collier = []
for yvar in yvars:
for xvar in xvars:
if xvar == yvar:
continue
tdf = df.filter(items=[yvar, xvar], axis=1)
tdf.replace([np.inf, -np.inf], np.nan, inplace=True)
tdf.dropna(how='any', inplace=True)
y_o = tdf[yvar].values.tolist()
x_o = tdf[xvar].values.tolist()
x_o, y_o = zip(*sorted(zip(x_o, y_o)))
x_o = np.array(x_o)
y_o = np.array(y_o)
#Create single dimension
x = x_o[:, np.newaxis]
y = y_o[:, np.newaxis]
inds = x.ravel().argsort() # Sort x values and get index
x = x.ravel()[inds].reshape(-1, 1)
y = y[inds] #Sort y according to x sorted index
for model_type in ['linear', 'quadratic', 'cubic']:
d = int()
if model_type == 'linear': d = 1
elif model_type == 'quadratic': d = 2
elif model_type == 'cubic': d = 3
if robust % 2 == 0:
# Run OLS
polynomial_features = PolynomialFeatures(degree = d)
xp = polynomial_features.fit_transform(x)
model = sm.OLS(y, xp).fit()
ypred = model.predict(xp)
ypred = ypred.tolist()
r2 = model.rsquared
try:
r2 = round(r2,4)
except:
pass
if r2 < 0:
r2 = 0
r2_adj = model.rsquared_adj
try:
r2_adj = round(r2_adj,4)
except:
pass
if r2_adj < 0:
r2_adj = 0
try:
y = y.flatten().tolist()
except:
pass
op_r2 = app_fxns.obs_pred_rsquare(np.array(y), np.array(ypred))
try:
op_r2 = round(op_r2, 4)
except:
pass
if op_r2 < 0:
op_r2 = 0
elif robust % 2 != 0:
# Rub Robust Regression
y_o = tdf[yvar].values.tolist()
x_o = tdf[xvar].values.tolist()
x_o, y_o = zip(*sorted(zip(x_o, y_o)))
x = np.array(x_o)
y = np.array(y_o)
# Create polynomial features up to 3rd degree
# Add a constant for the intercept
if d == 1:
# For a 1st-degree
X_poly = sm.add_constant(x)
elif d == 2:
# For a 2nd-degree polynomial (X, X^2)
X_poly = np.column_stack((x, x**2))
X_poly = sm.add_constant(X_poly)
elif d == 3:
# For a 3rd-degree polynomial (X, X^2, X^3)
X_poly = np.column_stack((x, x**2, x**3))
X_poly = sm.add_constant(X_poly)
# Fit a robust polynomial regression model
model = sm.RLM(y, X_poly, M=sm.robust.norms.TukeyBiweight(),
missing='drop').fit()
ypred = model.fittedvalues
op_r2 = round(app_fxns.obs_pred_rsquare(y, ypred),4)
if op_r2 < 0:
op_r2 = 0
r2 = 'N/A'
r2_adj = 'N/A'
# Shapiro-Wilk for normally distributed errors
#val, sw_p = stats.shapiro(model.resid)
#shapiro_wilk.append(sw_p)
# Jarque-Bera for normally distributed errors
jb_test = sms.jarque_bera(model.resid)
jb_p = round(jb_test[1], 4)
jarque_bera.append(jb_p)
# Durbin-Watson for autocorrelation
dw = stattools.durbin_watson(model.resid)
durbin_watson.append(round(dw, 4))
# Breusch-Pagan for heteroskedasticity
bp_test = sms.het_breuschpagan(model.resid, model.model.exog)
bp_p = round(bp_test[1],4)
breusch_pagan.append(bp_p)
# Harvey-Collier multiplier test for linearity
if d == 1:
try:
skip = 10 #len(model.params) # bug in linear_harvey_collier
rr = sms.recursive_olsresiduals(model, skip=skip,
alpha=0.95, order_by=None)
hc_test = stats.ttest_1samp(rr[3][skip:], 0)
hc_p = round(hc_test[1], 4)
except:
hc_p = 'Inconclusive'
else:
hc_p = 'N/A'
harvey_collier.append(hc_p)
intercepts.append(model.params[0])
coefs.append(model.params[1:])
poly_coefs = model.params[1:].tolist()
poly_coefs.reverse()
poly_exponents = list(range(1, len(poly_coefs)+1))
poly_exponents.reverse()
eqn = 'y = '
for i, p in enumerate(poly_coefs):
exp = poly_exponents[i]
if exp == 1:
exp = 'x'
elif exp == 2:
exp = 'x²'
elif exp == 3:
exp = 'x³'
if i == 0:
p = round(p, 4)
eqn = eqn + str(p) + exp
else:
if p >= 0:
p = round(p, 4)
eqn = eqn + ' + ' + str(p) + exp
else:
p = round(p, 4)
eqn = eqn + ' - ' + str(np.abs(p)) + exp
b = model.params[0]
if b >= 0:
b = round(b, 4)
eqn = eqn + ' + ' + str(b)
else:
b = round(b, 4)
eqn = eqn + ' - ' + str(np.abs(b))
eqns.append(eqn)
if robust % 2 == 0:
aic = round(model.aic,4)
bic = round(model.bic,4)
fp = round(model.f_pvalue,4)
llf = round(model.llf,4)
else:
aic = 'N/A'
bic = 'N/A'
fp = 'N/A'
llf = 'N/A'
Yvars.append(yvar)
Xvars.append(xvar)
yvar_short = str(yvar)
xvar_short = str(xvar)
if len(yvar) > 60:
yvar_short = yvar[0:30] + ' ... ' + yvar[-30:]
if len(xvar) > 60:
xvar_short = xvar[0:30] + ' ... ' + xvar[-30:]
Yvars_short.append(yvar_short)
Xvars_short.append(xvar_short)
models.append(model_type)
r2s.append(r2)
adj_r2s.append(r2_adj)
obs_pred_r2s.append(op_r2)
pvals.append(fp)
bics.append(bic)
aics.append(aic)
llf_ls.append(llf)
ns.append(len(x))
Xs.append(x_o)
Ys.append(y_o)
PredY.append(ypred)
del df
cols = ['y-variable', 'x-variable', 'Model', 'r-square', 'adj. r-square',
'obs vs. pred r-square', 'p-value', 'intercept', 'coefficients', 'AIC',
'BIC', 'log-likelihood', 'Durbin-Watson', 'Jarque-Bera (p-value)',
'Breusch-Pagan (p-value)', 'Harvey-Collier (p-value)']
df_models = pd.DataFrame(columns=cols)
df_models['y-variable'] = Yvars
df_models['x-variable'] = Xvars
df_models['y-variable (short)'] = Yvars_short
df_models['x-variable (short)'] = Xvars_short
df_models['Model'] = models
df_models['r-square'] = r2s
df_models['adj. r-square'] = adj_r2s
df_models['obs vs. pred r-square'] = obs_pred_r2s
df_models['p-value'] = pvals
df_models['AIC'] = aics
df_models['BIC'] = bics
df_models['log-likelihood'] = llf_ls
df_models['ys'] = Ys
df_models['xs'] = Xs
df_models['pred_y'] = PredY
df_models['intercept'] = intercepts
df_models['coefficients'] = coefs
df_models['equation'] = eqns
df_models['sample size'] = ns
df_models['Durbin-Watson'] = durbin_watson
df_models['Jarque-Bera (p-value)'] = jarque_bera
df_models['Breusch-Pagan (p-value)'] = breusch_pagan
df_models['Harvey-Collier (p-value)'] = harvey_collier
###################### Figure ########################
fig_data = []
df_models['label'] = df_models['y-variable (short)'] + '<br>' + ' vs.'
df_models['label'] = df_models['label'] + '<br>' + df_models['x-variable (short)']
tdf = df_models[df_models['Model'] == 'cubic']
tdf.sort_values(by='r-square', inplace=True, ascending=False)
tdf = tdf.head(10)
labels = tdf['label'].tolist()
del tdf
used = []
models = ['linear', 'quadratic', 'cubic']
for i, label in enumerate(labels):
if label in used:
continue
used.append(label)
clr = "#" + "%06x" % random.randint(0, 0xFFFFFF)
for model in models:
tdf = df_models[(df_models['label'] == label) & (df_models['Model'] == model)]
obs_y = tdf['ys'].tolist()
obs_y = obs_y[0]
obs_x = tdf['xs'].tolist()
obs_x = obs_x[0]
pred_y = tdf['pred_y'].tolist()
pred_y = pred_y[0]
r2 = tdf['r-square'].tolist()
r2 = r2[0]
op_r2 = tdf['obs vs. pred r-square'].tolist()
op_r2 = op_r2[0]
obs_x, obs_y, pred_y = zip(*sorted(zip(obs_x, obs_y, pred_y)))
if model == 'linear':
fig_data.append(
go.Scatter(
x = obs_x,
y = obs_y,
mode = "markers",
name = label,
opacity = 0.75,
marker = dict(size=10,
color=clr)
)
)
line_dict = dict(color = clr, width = 2)
if model == 'quadratic':
line_dict = dict(color = clr, width = 2, dash='dash')
elif model == 'cubic':
line_dict = dict(color = clr, width = 2, dash='dot')
if robust % 2 != 0:
r2 = round(op_r2, 4)
fig_data.append(
go.Scatter(
x = obs_x,
y = pred_y,
mode = "lines",
name = model + ': r<sup>2</sup> = '+str(r2),
opacity = 0.75,
line = line_dict,
)
)
figure = go.Figure(
data = fig_data,
layout = go.Layout(
xaxis = dict(
title = dict(
text = "<b>X</b>",
font = dict(
family = '"Open Sans", "HelveticaNeue", "Helvetica Neue",'
" Helvetica, Arial, sans-serif",
size = 18,
),
),
showticklabels = True,
),
yaxis = dict(
title = dict(
text = "<b>Y</b>",
font = dict(
family = '"Open Sans", "HelveticaNeue", "Helvetica Neue",'
" Helvetica, Arial, sans-serif",
size = 18,
),
),
showticklabels = True,
),
margin = dict(l=60, r=30, b=10, t=40),
showlegend = True,
height = 400,
paper_bgcolor = "rgb(245, 247, 249)",
plot_bgcolor = "rgb(245, 247, 249)",
),
)
################## Table ######################
df_models.drop(labels=['x-variable (short)', 'y-variable (short)'], axis=1, inplace=True)
df_models = df_models[df_models['x-variable'].isin(xvars)]
df_models = df_models[df_models['y-variable'].isin(yvars)]
cols = ['y-variable', 'x-variable', 'sample size', 'Model', 'r-square', 'adj. r-square',
'obs vs. pred r-square', 'p-value', 'AIC', 'BIC', 'log-likelihood',
'Durbin-Watson', 'Jarque-Bera (p-value)', 'Breusch-Pagan (p-value)',
'Harvey-Collier (p-value)', 'equation']
df_table = df_models.filter(items=cols)
df_table.sort_values(by='adj. r-square', inplace=True, ascending=False)
dashT = dash_table.DataTable(
data=df_table.to_dict('records'),
columns=[{'id': c, 'name': c} for c in df_table.columns],
export_format="csv",
page_action='none',
sort_action="native",
sort_mode="multi",
filter_action="native",
style_table={'height': '500px',
'overflowY': 'auto',
},
style_cell={'padding':'5px',
'minwidth':'140px',
'width':'160px',
'maxwidth':'160px',
'whiteSpace':'normal',
'textAlign': 'center',
},
)
del df_models
txt1 = "This figure displays up to 10 pairs of features sharing the strongest linear "
txt1 += "and curvilinear (polynomial) relationships. "
txt1 += "Polynomial regression is useful when relationships are noticeably curved. "
txt1 += "Quadratic models account for one curve and cubic models account for two. "
txt1 += "When interpreting performance, consider whether or not a curvier model "
txt1 += "produces meaningfully greater r\u00B2"
txt2 = "The Durbin-Watson statistic ranges between 0 and 4. The closer it is to 2, "
txt2 += "the more independent the observations. "
txt2 += "Significant Jarque-Bera tests (p < 0.05) indicate non-normality. "
txt2 += "Significant Breusch-Pagan tests (p < 0.05) indicate heteroskedasticity. "
txt2 += "Significant Harvey-Collier test (p < 0.05) indicate non-linearity. "
if robust == 1:
txt2 += "\nNote, outputs of robust regression do not include AIC, BIC, "
txt2 += "log-likelihood, or p-values from an F-test, or typical r-square and "
txt2 += "adjusted r-square values. Instead, r-square values for robust regression "
txt2 += "are based on observed vs predicted, i.e., a linear regression between "
txt2 += "observed and predicted values with the slope constrained to 1 and the "
txt2 += "intercept constrained to 0."
return figure, dashT, "", txt1, txt2