-
Notifications
You must be signed in to change notification settings - Fork 0
/
s2.html
583 lines (538 loc) · 40.1 KB
/
s2.html
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
<!DOCTYPE html>
<html lang="" xml:lang="">
<head>
<meta charset="utf-8" />
<meta http-equiv="X-UA-Compatible" content="IE=edge" />
<title>Chapter 5 Step 2: Propensity score Matching | Understanding Propensity Score Matching</title>
<meta name="description" content="Chapter 5 Step 2: Propensity score Matching | Understanding Propensity Score Matching." />
<meta name="generator" content="bookdown 0.28 and GitBook 2.6.7" />
<meta property="og:title" content="Chapter 5 Step 2: Propensity score Matching | Understanding Propensity Score Matching" />
<meta property="og:type" content="book" />
<meta property="og:description" content="Chapter 5 Step 2: Propensity score Matching | Understanding Propensity Score Matching." />
<meta name="github-repo" content="ehsanx/UnderstandingPropensityScore" />
<meta name="twitter:card" content="summary" />
<meta name="twitter:title" content="Chapter 5 Step 2: Propensity score Matching | Understanding Propensity Score Matching" />
<meta name="twitter:description" content="Chapter 5 Step 2: Propensity score Matching | Understanding Propensity Score Matching." />
<meta name="author" content="Ehsan Karim" />
<meta name="date" content="2023-03-19" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<meta name="apple-mobile-web-app-capable" content="yes" />
<meta name="apple-mobile-web-app-status-bar-style" content="black" />
<link rel="prev" href="s1.html"/>
<link rel="next" href="s3.html"/>
<script src="libs/jquery-3.6.0/jquery-3.6.0.min.js"></script>
<script src="https://cdn.jsdelivr.net/npm/fuse.js@6.4.6/dist/fuse.min.js"></script>
<link href="libs/gitbook-2.6.7/css/style.css" rel="stylesheet" />
<link href="libs/gitbook-2.6.7/css/plugin-table.css" rel="stylesheet" />
<link href="libs/gitbook-2.6.7/css/plugin-bookdown.css" rel="stylesheet" />
<link href="libs/gitbook-2.6.7/css/plugin-highlight.css" rel="stylesheet" />
<link href="libs/gitbook-2.6.7/css/plugin-search.css" rel="stylesheet" />
<link href="libs/gitbook-2.6.7/css/plugin-fontsettings.css" rel="stylesheet" />
<link href="libs/gitbook-2.6.7/css/plugin-clipboard.css" rel="stylesheet" />
<link href="libs/anchor-sections-1.1.0/anchor-sections.css" rel="stylesheet" />
<link href="libs/anchor-sections-1.1.0/anchor-sections-hash.css" rel="stylesheet" />
<script src="libs/anchor-sections-1.1.0/anchor-sections.js"></script>
<script src="libs/kePrint-0.0.1/kePrint.js"></script>
<link href="libs/lightable-0.0.1/lightable.css" rel="stylesheet" />
<script type="text/javascript">
// toggle visibility of R source blocks in R Markdown output
function toggle_R() {
var x = document.getElementsByClassName('r');
if (x.length == 0) return;
function toggle_vis(o) {
var d = o.style.display;
o.style.display = (d == 'block' || d == '') ? 'none':'block';
}
for (i = 0; i < x.length; i++) {
var y = x[i];
if (y.tagName.toLowerCase() === 'pre') toggle_vis(y);
}
var elem = document.getElementById("myButton1");
if (elem.value === "Hide Global") elem.value = "Show Global";
else elem.value = "Hide Global";
}
document.write('<input onclick="toggle_R();" type="button" value="Hide Global" id="myButton1" style="position: absolute; top: 10%; right: 2%; z-index: 200"></input>')
</script>
<style type="text/css">
pre > code.sourceCode { white-space: pre; position: relative; }
pre > code.sourceCode > span { display: inline-block; line-height: 1.25; }
pre > code.sourceCode > span:empty { height: 1.2em; }
.sourceCode { overflow: visible; }
code.sourceCode > span { color: inherit; text-decoration: inherit; }
pre.sourceCode { margin: 0; }
@media screen {
div.sourceCode { overflow: auto; }
}
@media print {
pre > code.sourceCode { white-space: pre-wrap; }
pre > code.sourceCode > span { text-indent: -5em; padding-left: 5em; }
}
pre.numberSource code
{ counter-reset: source-line 0; }
pre.numberSource code > span
{ position: relative; left: -4em; counter-increment: source-line; }
pre.numberSource code > span > a:first-child::before
{ content: counter(source-line);
position: relative; left: -1em; text-align: right; vertical-align: baseline;
border: none; display: inline-block;
-webkit-touch-callout: none; -webkit-user-select: none;
-khtml-user-select: none; -moz-user-select: none;
-ms-user-select: none; user-select: none;
padding: 0 4px; width: 4em;
color: #aaaaaa;
}
pre.numberSource { margin-left: 3em; border-left: 1px solid #aaaaaa; padding-left: 4px; }
div.sourceCode
{ }
@media screen {
pre > code.sourceCode > span > a:first-child::before { text-decoration: underline; }
}
code span.al { color: #ff0000; font-weight: bold; } /* Alert */
code span.an { color: #60a0b0; font-weight: bold; font-style: italic; } /* Annotation */
code span.at { color: #7d9029; } /* Attribute */
code span.bn { color: #40a070; } /* BaseN */
code span.bu { color: #008000; } /* BuiltIn */
code span.cf { color: #007020; font-weight: bold; } /* ControlFlow */
code span.ch { color: #4070a0; } /* Char */
code span.cn { color: #880000; } /* Constant */
code span.co { color: #60a0b0; font-style: italic; } /* Comment */
code span.cv { color: #60a0b0; font-weight: bold; font-style: italic; } /* CommentVar */
code span.do { color: #ba2121; font-style: italic; } /* Documentation */
code span.dt { color: #902000; } /* DataType */
code span.dv { color: #40a070; } /* DecVal */
code span.er { color: #ff0000; font-weight: bold; } /* Error */
code span.ex { } /* Extension */
code span.fl { color: #40a070; } /* Float */
code span.fu { color: #06287e; } /* Function */
code span.im { color: #008000; font-weight: bold; } /* Import */
code span.in { color: #60a0b0; font-weight: bold; font-style: italic; } /* Information */
code span.kw { color: #007020; font-weight: bold; } /* Keyword */
code span.op { color: #666666; } /* Operator */
code span.ot { color: #007020; } /* Other */
code span.pp { color: #bc7a00; } /* Preprocessor */
code span.sc { color: #4070a0; } /* SpecialChar */
code span.ss { color: #bb6688; } /* SpecialString */
code span.st { color: #4070a0; } /* String */
code span.va { color: #19177c; } /* Variable */
code span.vs { color: #4070a0; } /* VerbatimString */
code span.wa { color: #60a0b0; font-weight: bold; font-style: italic; } /* Warning */
</style>
<style type="text/css">
/* Used with Pandoc 2.11+ new --citeproc when CSL is used */
div.csl-bib-body { }
div.csl-entry {
clear: both;
}
.hanging div.csl-entry {
margin-left:2em;
text-indent:-2em;
}
div.csl-left-margin {
min-width:2em;
float:left;
}
div.csl-right-inline {
margin-left:2em;
padding-left:1em;
}
div.csl-indent {
margin-left: 2em;
}
</style>
<link rel="stylesheet" href="style.css" type="text/css" />
</head>
<body>
<div class="book without-animation with-summary font-size-2 font-family-1" data-basepath=".">
<div class="book-summary">
<nav role="navigation">
<ul class="summary">
<li><a href="./">Understanding Propensity Score Matching</a></li>
<li class="divider"></li>
<li class="chapter" data-level="" data-path="index.html"><a href="index.html"><i class="fa fa-check"></i>Preamble</a>
<ul>
<li class="chapter" data-level="" data-path="index.html"><a href="index.html#description"><i class="fa fa-check"></i>Description</a>
<ul>
<li class="chapter" data-level="" data-path="index.html"><a href="index.html#main-references"><i class="fa fa-check"></i>Main references</a></li>
<li class="chapter" data-level="" data-path="index.html"><a href="index.html#version-history"><i class="fa fa-check"></i>Version history</a></li>
</ul></li>
<li class="chapter" data-level="" data-path="index.html"><a href="index.html#prerequisites"><i class="fa fa-check"></i>Prerequisites</a>
<ul>
<li class="chapter" data-level="" data-path="index.html"><a href="index.html#license"><i class="fa fa-check"></i>License</a></li>
<li class="chapter" data-level="" data-path="index.html"><a href="index.html#comments"><i class="fa fa-check"></i>Comments</a></li>
</ul></li>
</ul></li>
<li class="chapter" data-level="1" data-path="terms.html"><a href="terms.html"><i class="fa fa-check"></i><b>1</b> Defining Parameter</a>
<ul>
<li class="chapter" data-level="1.1" data-path="terms.html"><a href="terms.html#epidemiological-research-goals"><i class="fa fa-check"></i><b>1.1</b> Epidemiological research goals</a></li>
<li class="chapter" data-level="1.2" data-path="terms.html"><a href="terms.html#potential-outcome"><i class="fa fa-check"></i><b>1.2</b> Potential outcome</a></li>
<li class="chapter" data-level="1.3" data-path="terms.html"><a href="terms.html#parameters-of-interest"><i class="fa fa-check"></i><b>1.3</b> Parameters of interest</a>
<ul>
<li class="chapter" data-level="1.3.1" data-path="terms.html"><a href="terms.html#te"><i class="fa fa-check"></i><b>1.3.1</b> TE</a></li>
<li class="chapter" data-level="1.3.2" data-path="terms.html"><a href="terms.html#ate"><i class="fa fa-check"></i><b>1.3.2</b> ATE</a></li>
<li class="chapter" data-level="1.3.3" data-path="terms.html"><a href="terms.html#interpretation-of-ate"><i class="fa fa-check"></i><b>1.3.3</b> Interpretation of ATE</a></li>
<li class="chapter" data-level="1.3.4" data-path="terms.html"><a href="terms.html#identifiability-assumptions"><i class="fa fa-check"></i><b>1.3.4</b> Identifiability Assumptions</a></li>
<li class="chapter" data-level="1.3.5" data-path="terms.html"><a href="terms.html#att"><i class="fa fa-check"></i><b>1.3.5</b> ATT</a></li>
<li class="chapter" data-level="1.3.6" data-path="terms.html"><a href="terms.html#interpretation-of-att"><i class="fa fa-check"></i><b>1.3.6</b> Interpretation of ATT</a></li>
<li class="chapter" data-level="1.3.7" data-path="terms.html"><a href="terms.html#att-vs.-ate"><i class="fa fa-check"></i><b>1.3.7</b> ATT vs. ATE</a></li>
</ul></li>
</ul></li>
<li class="chapter" data-level="2" data-path="balance.html"><a href="balance.html"><i class="fa fa-check"></i><b>2</b> Balance and Overlap</a>
<ul>
<li class="chapter" data-level="2.1" data-path="balance.html"><a href="balance.html#balance-1"><i class="fa fa-check"></i><b>2.1</b> Balance</a>
<ul>
<li class="chapter" data-level="2.1.1" data-path="balance.html"><a href="balance.html#measures-of-balance"><i class="fa fa-check"></i><b>2.1.1</b> Measures of Balance</a></li>
</ul></li>
<li class="chapter" data-level="2.2" data-path="balance.html"><a href="balance.html#adjustment"><i class="fa fa-check"></i><b>2.2</b> Adjustment</a>
<ul>
<li class="chapter" data-level="2.2.1" data-path="balance.html"><a href="balance.html#why-adjust"><i class="fa fa-check"></i><b>2.2.1</b> Why adjust?</a></li>
<li class="chapter" data-level="2.2.2" data-path="balance.html"><a href="balance.html#adjustment-methods"><i class="fa fa-check"></i><b>2.2.2</b> Adjustment Methods</a></li>
</ul></li>
<li class="chapter" data-level="2.3" data-path="balance.html"><a href="balance.html#lack-of-overlap"><i class="fa fa-check"></i><b>2.3</b> Lack of overlap</a></li>
</ul></li>
<li class="chapter" data-level="3" data-path="ps.html"><a href="ps.html"><i class="fa fa-check"></i><b>3</b> Propensity score</a>
<ul>
<li class="chapter" data-level="3.1" data-path="ps.html"><a href="ps.html#motivating-problem"><i class="fa fa-check"></i><b>3.1</b> Motivating problem</a></li>
<li class="chapter" data-level="3.2" data-path="ps.html"><a href="ps.html#defining-propensity-score"><i class="fa fa-check"></i><b>3.2</b> Defining Propensity score</a>
<ul>
<li class="chapter" data-level="3.2.1" data-path="ps.html"><a href="ps.html#theoretical-result"><i class="fa fa-check"></i><b>3.2.1</b> Theoretical result</a></li>
<li class="chapter" data-level="3.2.2" data-path="ps.html"><a href="ps.html#assumptions"><i class="fa fa-check"></i><b>3.2.2</b> Assumptions</a></li>
<li class="chapter" data-level="3.2.3" data-path="ps.html"><a href="ps.html#ways-to-use-ps"><i class="fa fa-check"></i><b>3.2.3</b> Ways to use PS</a></li>
</ul></li>
<li class="chapter" data-level="3.3" data-path="ps.html"><a href="ps.html#ps-matching-steps"><i class="fa fa-check"></i><b>3.3</b> PS Matching Steps</a></li>
</ul></li>
<li class="chapter" data-level="4" data-path="s1.html"><a href="s1.html"><i class="fa fa-check"></i><b>4</b> Step 1: Exposure modelling</a>
<ul>
<li class="chapter" data-level="4.1" data-path="s1.html"><a href="s1.html#model-specification"><i class="fa fa-check"></i><b>4.1</b> Model specification</a>
<ul>
<li class="chapter" data-level="4.1.1" data-path="s1.html"><a href="s1.html#updating-model-specification"><i class="fa fa-check"></i><b>4.1.1</b> Updating model specification</a></li>
<li class="chapter" data-level="4.1.2" data-path="s1.html"><a href="s1.html#stability-of-ps"><i class="fa fa-check"></i><b>4.1.2</b> Stability of PS</a></li>
</ul></li>
<li class="chapter" data-level="4.2" data-path="s1.html"><a href="s1.html#variables-to-adjust"><i class="fa fa-check"></i><b>4.2</b> Variables to adjust</a>
<ul>
<li class="chapter" data-level="4.2.1" data-path="s1.html"><a href="s1.html#best-approach"><i class="fa fa-check"></i><b>4.2.1</b> Best approach</a></li>
<li class="chapter" data-level="4.2.2" data-path="s1.html"><a href="s1.html#general-guideline-of-type-of-variables"><i class="fa fa-check"></i><b>4.2.2</b> General guideline of type of variables</a></li>
<li class="chapter" data-level="4.2.3" data-path="s1.html"><a href="s1.html#what-not-to-include"><i class="fa fa-check"></i><b>4.2.3</b> What NOT to include</a></li>
<li class="chapter" data-level="4.2.4" data-path="s1.html"><a href="s1.html#mediators"><i class="fa fa-check"></i><b>4.2.4</b> Mediators</a></li>
<li class="chapter" data-level="4.2.5" data-path="s1.html"><a href="s1.html#unmeasured-confounding"><i class="fa fa-check"></i><b>4.2.5</b> Unmeasured confounding</a></li>
</ul></li>
<li class="chapter" data-level="4.3" data-path="s1.html"><a href="s1.html#model-selection"><i class="fa fa-check"></i><b>4.3</b> Model selection</a>
<ul>
<li class="chapter" data-level="4.3.1" data-path="s1.html"><a href="s1.html#based-on-association-with-outcome"><i class="fa fa-check"></i><b>4.3.1</b> Based on association with outcome</a></li>
<li class="chapter" data-level="4.3.2" data-path="s1.html"><a href="s1.html#based-on-association-with-exposure"><i class="fa fa-check"></i><b>4.3.2</b> Based on association with exposure</a></li>
</ul></li>
<li class="chapter" data-level="4.4" data-path="s1.html"><a href="s1.html#alternative-modelling-strategies"><i class="fa fa-check"></i><b>4.4</b> Alternative modelling strategies</a></li>
<li class="chapter" data-level="4.5" data-path="s1.html"><a href="s1.html#ps-estimation"><i class="fa fa-check"></i><b>4.5</b> PS estimation</a></li>
</ul></li>
<li class="chapter" data-level="5" data-path="s2.html"><a href="s2.html"><i class="fa fa-check"></i><b>5</b> Step 2: Propensity score Matching</a>
<ul>
<li class="chapter" data-level="5.1" data-path="s2.html"><a href="s2.html#matching-method-nn"><i class="fa fa-check"></i><b>5.1</b> Matching method NN</a></li>
<li class="chapter" data-level="5.2" data-path="s2.html"><a href="s2.html#initial-fit"><i class="fa fa-check"></i><b>5.2</b> Initial fit</a></li>
<li class="chapter" data-level="5.3" data-path="s2.html"><a href="s2.html#fine-tuning-add-caliper"><i class="fa fa-check"></i><b>5.3</b> Fine tuning: add caliper</a></li>
<li class="chapter" data-level="5.4" data-path="s2.html"><a href="s2.html#things-to-keep-track-of"><i class="fa fa-check"></i><b>5.4</b> Things to keep track of</a></li>
<li class="chapter" data-level="5.5" data-path="s2.html"><a href="s2.html#matches"><i class="fa fa-check"></i><b>5.5</b> Matches</a></li>
<li class="chapter" data-level="5.6" data-path="s2.html"><a href="s2.html#other-matching-algorithms"><i class="fa fa-check"></i><b>5.6</b> Other matching algorithms</a></li>
</ul></li>
<li class="chapter" data-level="6" data-path="s3.html"><a href="s3.html"><i class="fa fa-check"></i><b>6</b> Step 3: Balance and overlap</a>
<ul>
<li class="chapter" data-level="6.1" data-path="s3.html"><a href="s3.html#assessment-of-balance-by-smd"><i class="fa fa-check"></i><b>6.1</b> Assessment of Balance by SMD</a></li>
<li class="chapter" data-level="6.2" data-path="s3.html"><a href="s3.html#smd-vs.-p-values"><i class="fa fa-check"></i><b>6.2</b> SMD vs. P-values</a></li>
<li class="chapter" data-level="6.3" data-path="s3.html"><a href="s3.html#vizualization-for-overlap"><i class="fa fa-check"></i><b>6.3</b> Vizualization for Overlap</a></li>
<li class="chapter" data-level="6.4" data-path="s3.html"><a href="s3.html#variance-ratio-1"><i class="fa fa-check"></i><b>6.4</b> Variance ratio</a></li>
<li class="chapter" data-level="6.5" data-path="s3.html"><a href="s3.html#close-inspection-of-boundaries"><i class="fa fa-check"></i><b>6.5</b> Close inspection of boundaries</a></li>
<li class="chapter" data-level="6.6" data-path="s3.html"><a href="s3.html#unsatirfactory-balance"><i class="fa fa-check"></i><b>6.6</b> Unsatirfactory balance</a></li>
</ul></li>
<li class="chapter" data-level="7" data-path="s4.html"><a href="s4.html"><i class="fa fa-check"></i><b>7</b> Step 4: Outcome modelling</a>
<ul>
<li class="chapter" data-level="7.1" data-path="s4.html"><a href="s4.html#crude-outcome-model"><i class="fa fa-check"></i><b>7.1</b> Crude outcome model</a></li>
<li class="chapter" data-level="7.2" data-path="s4.html"><a href="s4.html#double-adjustment"><i class="fa fa-check"></i><b>7.2</b> Double-adjustment</a></li>
<li class="chapter" data-level="7.3" data-path="s4.html"><a href="s4.html#adjusted-outcome-model"><i class="fa fa-check"></i><b>7.3</b> Adjusted outcome model</a></li>
<li class="chapter" data-level="7.4" data-path="s4.html"><a href="s4.html#variance-considerations"><i class="fa fa-check"></i><b>7.4</b> Variance considerations</a>
<ul>
<li class="chapter" data-level="7.4.1" data-path="s4.html"><a href="s4.html#cluster-option"><i class="fa fa-check"></i><b>7.4.1</b> Cluster option</a></li>
<li class="chapter" data-level="7.4.2" data-path="s4.html"><a href="s4.html#bootstrap"><i class="fa fa-check"></i><b>7.4.2</b> Bootstrap</a></li>
</ul></li>
<li class="chapter" data-level="7.5" data-path="s4.html"><a href="s4.html#estimate-obtained"><i class="fa fa-check"></i><b>7.5</b> Estimate obtained</a></li>
</ul></li>
<li class="chapter" data-level="8" data-path="compare.html"><a href="compare.html"><i class="fa fa-check"></i><b>8</b> PS vs. Regression</a>
<ul>
<li class="chapter" data-level="8.1" data-path="compare.html"><a href="compare.html#data-simulation"><i class="fa fa-check"></i><b>8.1</b> Data Simulation</a></li>
<li class="chapter" data-level="8.2" data-path="compare.html"><a href="compare.html#treatment-effect-from-counterfactuals"><i class="fa fa-check"></i><b>8.2</b> Treatment effect from counterfactuals</a></li>
<li class="chapter" data-level="8.3" data-path="compare.html"><a href="compare.html#treatment-effect-from-regression"><i class="fa fa-check"></i><b>8.3</b> Treatment effect from Regression</a></li>
<li class="chapter" data-level="8.4" data-path="compare.html"><a href="compare.html#treatment-effect-from-ps"><i class="fa fa-check"></i><b>8.4</b> Treatment effect from PS</a></li>
<li class="chapter" data-level="8.5" data-path="compare.html"><a href="compare.html#non-linear-model"><i class="fa fa-check"></i><b>8.5</b> Non-linear Model</a>
<ul>
<li class="chapter" data-level="8.5.1" data-path="compare.html"><a href="compare.html#data-generation"><i class="fa fa-check"></i><b>8.5.1</b> Data generation</a></li>
<li class="chapter" data-level="8.5.2" data-path="compare.html"><a href="compare.html#regression"><i class="fa fa-check"></i><b>8.5.2</b> Regression</a></li>
<li class="chapter" data-level="8.5.3" data-path="compare.html"><a href="compare.html#ps-1"><i class="fa fa-check"></i><b>8.5.3</b> PS</a></li>
<li class="chapter" data-level="8.5.4" data-path="compare.html"><a href="compare.html#machine-learning"><i class="fa fa-check"></i><b>8.5.4</b> Machine learning</a></li>
<li class="chapter" data-level="8.5.5" data-path="compare.html"><a href="compare.html#regression-is-doomed"><i class="fa fa-check"></i><b>8.5.5</b> Regression is doomed?</a></li>
</ul></li>
</ul></li>
<li class="chapter" data-level="9" data-path="misspecify.html"><a href="misspecify.html"><i class="fa fa-check"></i><b>9</b> PS vs. Double robust methods</a>
<ul>
<li class="chapter" data-level="9.1" data-path="misspecify.html"><a href="misspecify.html#complex-data-simulation"><i class="fa fa-check"></i><b>9.1</b> Complex Data Simulation</a>
<ul>
<li class="chapter" data-level="" data-path="misspecify.html"><a href="misspecify.html#true-exposure-model"><i class="fa fa-check"></i>True Exposure Model</a></li>
<li class="chapter" data-level="" data-path="misspecify.html"><a href="misspecify.html#true-outcome-model"><i class="fa fa-check"></i>True Outcome Model</a></li>
<li class="chapter" data-level="" data-path="misspecify.html"><a href="misspecify.html#outcomes-and-exposures-are-complex-functions-of-measured-covariates"><i class="fa fa-check"></i>Outcomes and exposures are complex functions of measured covariates</a></li>
</ul></li>
<li class="chapter" data-level="9.2" data-path="misspecify.html"><a href="misspecify.html#understanding-finite-sample-bias"><i class="fa fa-check"></i><b>9.2</b> Understanding finite sample bias</a></li>
<li class="chapter" data-level="9.3" data-path="misspecify.html"><a href="misspecify.html#estimation-using-different-methods"><i class="fa fa-check"></i><b>9.3</b> Estimation using different methods</a>
<ul>
<li class="chapter" data-level="9.3.1" data-path="misspecify.html"><a href="misspecify.html#regression-1"><i class="fa fa-check"></i><b>9.3.1</b> Regression</a></li>
<li class="chapter" data-level="9.3.2" data-path="misspecify.html"><a href="misspecify.html#propensity-score"><i class="fa fa-check"></i><b>9.3.2</b> Propensity score</a></li>
<li class="chapter" data-level="9.3.3" data-path="misspecify.html"><a href="misspecify.html#double-machine-learning-method"><i class="fa fa-check"></i><b>9.3.3</b> Double machine learning method</a></li>
<li class="chapter" data-level="9.3.4" data-path="misspecify.html"><a href="misspecify.html#augmented-inverse-probability-weighting"><i class="fa fa-check"></i><b>9.3.4</b> Augmented Inverse probability weighting</a></li>
<li class="chapter" data-level="9.3.5" data-path="misspecify.html"><a href="misspecify.html#double-robust-method-tmle"><i class="fa fa-check"></i><b>9.3.5</b> Double robust method (TMLE)</a></li>
</ul></li>
</ul></li>
<li class="chapter" data-level="10" data-path="guide.html"><a href="guide.html"><i class="fa fa-check"></i><b>10</b> Reporting Guidelines</a>
<ul>
<li class="chapter" data-level="10.1" data-path="guide.html"><a href="guide.html#discipline-specific-reviews"><i class="fa fa-check"></i><b>10.1</b> Discipline-specific Reviews</a></li>
<li class="chapter" data-level="10.2" data-path="guide.html"><a href="guide.html#suggested-guidelines"><i class="fa fa-check"></i><b>10.2</b> Suggested Guidelines</a></li>
<li class="chapter" data-level="10.3" data-path="guide.html"><a href="guide.html#additional-topics"><i class="fa fa-check"></i><b>10.3</b> Additional topics</a></li>
</ul></li>
<li class="chapter" data-level="11" data-path="final.html"><a href="final.html"><i class="fa fa-check"></i><b>11</b> Final Words</a>
<ul>
<li class="chapter" data-level="11.1" data-path="final.html"><a href="final.html#common-misconception"><i class="fa fa-check"></i><b>11.1</b> Common misconception</a></li>
<li class="chapter" data-level="11.2" data-path="final.html"><a href="final.html#benifits-of-ps"><i class="fa fa-check"></i><b>11.2</b> Benifits of PS</a></li>
<li class="chapter" data-level="11.3" data-path="final.html"><a href="final.html#limitations-of-ps"><i class="fa fa-check"></i><b>11.3</b> Limitations of PS</a></li>
<li class="chapter" data-level="11.4" data-path="final.html"><a href="final.html#when-ps-may-not-be-useful"><i class="fa fa-check"></i><b>11.4</b> When PS may not be useful?</a></li>
<li class="chapter" data-level="11.5" data-path="final.html"><a href="final.html#software"><i class="fa fa-check"></i><b>11.5</b> Software</a></li>
<li class="chapter" data-level="11.6" data-path="final.html"><a href="final.html#further-resources"><i class="fa fa-check"></i><b>11.6</b> Further Resources</a></li>
</ul></li>
<li class="chapter" data-level="" data-path="references.html"><a href="references.html"><i class="fa fa-check"></i>References</a></li>
<li class="divider"></li>
<li><a href="https://ehsank.com/" target="blank">Ehsan Karim</a></li>
</ul>
</nav>
</div>
<div class="book-body">
<div class="body-inner">
<div class="book-header" role="navigation">
<h1>
<i class="fa fa-circle-o-notch fa-spin"></i><a href="./">Understanding Propensity Score Matching</a>
</h1>
</div>
<div class="page-wrapper" tabindex="-1" role="main">
<div class="page-inner">
<section class="normal" id="section-">
<div id="s2" class="section level1 hasAnchor" number="5">
<h1><span class="header-section-number">Chapter 5</span> Step 2: Propensity score Matching<a href="s2.html#s2" class="anchor-section" aria-label="Anchor link to header"></a></h1>
<ul>
<li>PS is a continuous variable.
<ul>
<li>Exact matching is not feasible.</li>
<li>Below is an example of control patient (treatment = 0) with PS = 0.25</li>
<li>We want to find a treated patient (treatment = 1) with PS closest to 0.25.</li>
</ul></li>
</ul>
<div class="sourceCode" id="cb37"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb37-1"><a href="s2.html#cb37-1" aria-hidden="true" tabindex="-1"></a><span class="fu">require</span>(cobalt)</span>
<span id="cb37-2"><a href="s2.html#cb37-2" aria-hidden="true" tabindex="-1"></a><span class="fu">library</span>(ggplot2)</span>
<span id="cb37-3"><a href="s2.html#cb37-3" aria-hidden="true" tabindex="-1"></a><span class="fu">bal.plot</span>(analytic, <span class="at">var.name =</span> <span class="st">"PS"</span>, </span>
<span id="cb37-4"><a href="s2.html#cb37-4" aria-hidden="true" tabindex="-1"></a> <span class="at">treat =</span> <span class="st">"diabetes"</span>, </span>
<span id="cb37-5"><a href="s2.html#cb37-5" aria-hidden="true" tabindex="-1"></a> <span class="at">which =</span> <span class="st">"both"</span>, </span>
<span id="cb37-6"><a href="s2.html#cb37-6" aria-hidden="true" tabindex="-1"></a> <span class="at">data =</span> analytic) <span class="sc">+</span> </span>
<span id="cb37-7"><a href="s2.html#cb37-7" aria-hidden="true" tabindex="-1"></a> <span class="fu">geom_vline</span>(<span class="at">xintercept=</span><span class="fl">0.25</span>, <span class="at">linetype=</span><span class="st">"dashed"</span>, <span class="at">color =</span> <span class="st">"red"</span>)</span></code></pre></div>
<p><img src="UnderstandingPropensityScore_files/figure-html/ps2cc444-1.png" width="672" /></p>
<div id="matching-method-nn" class="section level2 hasAnchor" number="5.1">
<h2><span class="header-section-number">5.1</span> Matching method NN<a href="s2.html#matching-method-nn" class="anchor-section" aria-label="Anchor link to header"></a></h2>
<p>Match using estimates propensity scores with the following choices (simplest choices)</p>
<ul>
<li><strong>Matching method</strong>:
<ul>
<li>nearest-neighbor (NN) matching</li>
</ul></li>
<li><strong>Can the same subject be chosen only once?</strong>:
<ul>
<li>matching without replacement</li>
</ul></li>
<li><strong>Closeness of the treated-untreated subjects</strong>:
<ul>
<li>with caliper = .2*SD of logit of propensity score</li>
</ul></li>
<li><strong>Ratio of treated-untreated subjects</strong>:
<ul>
<li>with 1:1 ratio (pair-matching)</li>
</ul></li>
</ul>
<p><img src="images/nn.png" width="50%" /></p>
</div>
<div id="initial-fit" class="section level2 hasAnchor" number="5.2">
<h2><span class="header-section-number">5.2</span> Initial fit<a href="s2.html#initial-fit" class="anchor-section" aria-label="Anchor link to header"></a></h2>
<p>1:1 NN Match using estimates propensity scores</p>
<div class="sourceCode" id="cb38"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb38-1"><a href="s2.html#cb38-1" aria-hidden="true" tabindex="-1"></a><span class="fu">set.seed</span>(<span class="dv">123</span>)</span>
<span id="cb38-2"><a href="s2.html#cb38-2" aria-hidden="true" tabindex="-1"></a><span class="fu">require</span>(MatchIt)</span>
<span id="cb38-3"><a href="s2.html#cb38-3" aria-hidden="true" tabindex="-1"></a>match.obj <span class="ot"><-</span> <span class="fu">matchit</span>(ps.formula, <span class="at">data =</span> analytic,</span>
<span id="cb38-4"><a href="s2.html#cb38-4" aria-hidden="true" tabindex="-1"></a> <span class="at">distance =</span> <span class="st">'logit'</span>, </span>
<span id="cb38-5"><a href="s2.html#cb38-5" aria-hidden="true" tabindex="-1"></a> <span class="at">method =</span> <span class="st">"nearest"</span>, </span>
<span id="cb38-6"><a href="s2.html#cb38-6" aria-hidden="true" tabindex="-1"></a> <span class="at">replace=</span><span class="cn">FALSE</span>,</span>
<span id="cb38-7"><a href="s2.html#cb38-7" aria-hidden="true" tabindex="-1"></a> <span class="at">ratio =</span> <span class="dv">1</span>)</span>
<span id="cb38-8"><a href="s2.html#cb38-8" aria-hidden="true" tabindex="-1"></a>analytic<span class="sc">$</span>PS <span class="ot"><-</span> match.obj<span class="sc">$</span>distance</span>
<span id="cb38-9"><a href="s2.html#cb38-9" aria-hidden="true" tabindex="-1"></a><span class="fu">summary</span>(match.obj<span class="sc">$</span>distance)</span></code></pre></div>
<pre><code>## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.003916 0.068128 0.169946 0.211268 0.312987 0.925132</code></pre>
<div class="sourceCode" id="cb40"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb40-1"><a href="s2.html#cb40-1" aria-hidden="true" tabindex="-1"></a>match.obj</span></code></pre></div>
<pre><code>## A matchit object
## - method: 1:1 nearest neighbor matching without replacement
## - distance: Propensity score
## - estimated with logistic regression
## - number of obs.: 1562 (original), 660 (matched)
## - target estimand: ATT
## - covariates: gender, age, race, education, married, bmi</code></pre>
</div>
<div id="fine-tuning-add-caliper" class="section level2 hasAnchor" number="5.3">
<h2><span class="header-section-number">5.3</span> Fine tuning: add caliper<a href="s2.html#fine-tuning-add-caliper" class="anchor-section" aria-label="Anchor link to header"></a></h2>
<p>2 SD of logit of the propensity score is suggested as a caliper to allow better comparability of the groups.</p>
<div class="sourceCode" id="cb42"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb42-1"><a href="s2.html#cb42-1" aria-hidden="true" tabindex="-1"></a>logitPS <span class="ot"><-</span> <span class="sc">-</span><span class="fu">log</span>(<span class="dv">1</span><span class="sc">/</span>analytic<span class="sc">$</span>PS <span class="sc">-</span> <span class="dv">1</span>) </span>
<span id="cb42-2"><a href="s2.html#cb42-2" aria-hidden="true" tabindex="-1"></a><span class="co"># logit of the propensity score</span></span>
<span id="cb42-3"><a href="s2.html#cb42-3" aria-hidden="true" tabindex="-1"></a>.<span class="dv">2</span><span class="sc">*</span><span class="fu">sd</span>(logitPS) <span class="co"># suggested in the literature</span></span></code></pre></div>
<pre><code>## [1] 0.2606266</code></pre>
<div class="sourceCode" id="cb44"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb44-1"><a href="s2.html#cb44-1" aria-hidden="true" tabindex="-1"></a><span class="co"># choosing too strict PS has unintended consequences </span></span></code></pre></div>
<div class="sourceCode" id="cb45"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb45-1"><a href="s2.html#cb45-1" aria-hidden="true" tabindex="-1"></a><span class="fu">set.seed</span>(<span class="dv">123</span>)</span>
<span id="cb45-2"><a href="s2.html#cb45-2" aria-hidden="true" tabindex="-1"></a><span class="fu">require</span>(MatchIt)</span>
<span id="cb45-3"><a href="s2.html#cb45-3" aria-hidden="true" tabindex="-1"></a>match.obj <span class="ot"><-</span> <span class="fu">matchit</span>(ps.formula, <span class="at">data =</span> analytic,</span>
<span id="cb45-4"><a href="s2.html#cb45-4" aria-hidden="true" tabindex="-1"></a> <span class="at">distance =</span> <span class="st">'logit'</span>, </span>
<span id="cb45-5"><a href="s2.html#cb45-5" aria-hidden="true" tabindex="-1"></a> <span class="at">method =</span> <span class="st">"nearest"</span>, </span>
<span id="cb45-6"><a href="s2.html#cb45-6" aria-hidden="true" tabindex="-1"></a> <span class="at">replace=</span><span class="cn">FALSE</span>,</span>
<span id="cb45-7"><a href="s2.html#cb45-7" aria-hidden="true" tabindex="-1"></a> <span class="at">caliper =</span> .<span class="dv">2</span><span class="sc">*</span><span class="fu">sd</span>(logitPS), </span>
<span id="cb45-8"><a href="s2.html#cb45-8" aria-hidden="true" tabindex="-1"></a> <span class="at">ratio =</span> <span class="dv">1</span>)</span>
<span id="cb45-9"><a href="s2.html#cb45-9" aria-hidden="true" tabindex="-1"></a>analytic<span class="sc">$</span>PS <span class="ot"><-</span> match.obj<span class="sc">$</span>distance</span>
<span id="cb45-10"><a href="s2.html#cb45-10" aria-hidden="true" tabindex="-1"></a><span class="fu">summary</span>(match.obj<span class="sc">$</span>distance)</span></code></pre></div>
<pre><code>## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.003916 0.068128 0.169946 0.211268 0.312987 0.925132</code></pre>
<div class="sourceCode" id="cb47"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb47-1"><a href="s2.html#cb47-1" aria-hidden="true" tabindex="-1"></a>match.obj</span></code></pre></div>
<pre><code>## A matchit object
## - method: 1:1 nearest neighbor matching without replacement
## - distance: Propensity score [caliper]
## - estimated with logistic regression
## - caliper: <distance> (0.045)
## - number of obs.: 1562 (original), 632 (matched)
## - target estimand: ATT
## - covariates: gender, age, race, education, married, bmi</code></pre>
</div>
<div id="things-to-keep-track-of" class="section level2 hasAnchor" number="5.4">
<h2><span class="header-section-number">5.4</span> Things to keep track of<a href="s2.html#things-to-keep-track-of" class="anchor-section" aria-label="Anchor link to header"></a></h2>
<ul>
<li>original sample size</li>
<li>matched sample size</li>
<li>percent reduction in sample</li>
<li>how many matched sets</li>
<li>some can be discarded because of no match; whether some sets are unequal</li>
</ul>
</div>
<div id="matches" class="section level2 hasAnchor" number="5.5">
<h2><span class="header-section-number">5.5</span> Matches<a href="s2.html#matches" class="anchor-section" aria-label="Anchor link to header"></a></h2>
<p>Taking a closer look at the matches</p>
<div class="sourceCode" id="cb49"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb49-1"><a href="s2.html#cb49-1" aria-hidden="true" tabindex="-1"></a><span class="co"># Ref: https://lists.gking.harvard.edu/pipermail/matchit/2013-October/000559.html</span></span>
<span id="cb49-2"><a href="s2.html#cb49-2" aria-hidden="true" tabindex="-1"></a>matches <span class="ot"><-</span> <span class="fu">as.data.frame</span>(match.obj<span class="sc">$</span>match.matrix)</span>
<span id="cb49-3"><a href="s2.html#cb49-3" aria-hidden="true" tabindex="-1"></a><span class="fu">colnames</span>(matches)<span class="ot"><-</span><span class="fu">c</span>(<span class="st">"matched_unit"</span>)</span>
<span id="cb49-4"><a href="s2.html#cb49-4" aria-hidden="true" tabindex="-1"></a>matches<span class="sc">$</span>matched_unit<span class="ot"><-</span><span class="fu">as.numeric</span>(</span>
<span id="cb49-5"><a href="s2.html#cb49-5" aria-hidden="true" tabindex="-1"></a> <span class="fu">as.character</span>(matches<span class="sc">$</span>matched_unit))</span>
<span id="cb49-6"><a href="s2.html#cb49-6" aria-hidden="true" tabindex="-1"></a>matches<span class="sc">$</span>treated_unit<span class="ot"><-</span><span class="fu">as.numeric</span>(<span class="fu">rownames</span>(matches))</span>
<span id="cb49-7"><a href="s2.html#cb49-7" aria-hidden="true" tabindex="-1"></a>matches.only<span class="ot"><-</span>matches[<span class="sc">!</span><span class="fu">is.na</span>(matches<span class="sc">$</span>matched_unit),]</span>
<span id="cb49-8"><a href="s2.html#cb49-8" aria-hidden="true" tabindex="-1"></a><span class="fu">head</span>(matches.only)</span></code></pre></div>
<pre><code>## matched_unit treated_unit
## 40 8496 40
## 56 3139 56
## 65 4192 65
## 66 94 66
## 86 2212 86
## 110 7154 110</code></pre>
<div class="sourceCode" id="cb51"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb51-1"><a href="s2.html#cb51-1" aria-hidden="true" tabindex="-1"></a>matched.data <span class="ot"><-</span> <span class="fu">match.data</span>(match.obj)</span>
<span id="cb51-2"><a href="s2.html#cb51-2" aria-hidden="true" tabindex="-1"></a><span class="fu">head</span>(<span class="fu">table</span>(matched.data<span class="sc">$</span>subclass))</span></code></pre></div>
<pre><code>##
## 1 2 3 4 5 6
## 2 2 2 2 2 2</code></pre>
<div class="sourceCode" id="cb53"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb53-1"><a href="s2.html#cb53-1" aria-hidden="true" tabindex="-1"></a><span class="fu">length</span>(<span class="fu">table</span>(matched.data<span class="sc">$</span>subclass))</span></code></pre></div>
<pre><code>## [1] 316</code></pre>
</div>
<div id="other-matching-algorithms" class="section level2 hasAnchor" number="5.6">
<h2><span class="header-section-number">5.6</span> Other matching algorithms<a href="s2.html#other-matching-algorithms" class="anchor-section" aria-label="Anchor link to header"></a></h2>
<ul>
<li>Variable NN ratio (e.g., 1:10) is usually better.</li>
<li>matching with replacement also possible</li>
</ul>
<p>But creates issue when calculating variances (but can be easily handled by controlling for ‘matching weights’).</p>
<p>Other possibilities</p>
<ul>
<li>Optimal</li>
<li>genetic matching</li>
<li>CEM</li>
</ul>
</div>
</div>
</section>
</div>
</div>
</div>
<a href="s1.html" class="navigation navigation-prev " aria-label="Previous page"><i class="fa fa-angle-left"></i></a>
<a href="s3.html" class="navigation navigation-next " aria-label="Next page"><i class="fa fa-angle-right"></i></a>
</div>
</div>
<script src="libs/gitbook-2.6.7/js/app.min.js"></script>
<script src="libs/gitbook-2.6.7/js/clipboard.min.js"></script>
<script src="libs/gitbook-2.6.7/js/plugin-search.js"></script>
<script src="libs/gitbook-2.6.7/js/plugin-sharing.js"></script>
<script src="libs/gitbook-2.6.7/js/plugin-fontsettings.js"></script>
<script src="libs/gitbook-2.6.7/js/plugin-bookdown.js"></script>
<script src="libs/gitbook-2.6.7/js/jquery.highlight.js"></script>
<script src="libs/gitbook-2.6.7/js/plugin-clipboard.js"></script>
<script>
gitbook.require(["gitbook"], function(gitbook) {
gitbook.start({
"sharing": {
"github": false,
"facebook": true,
"twitter": true,
"linkedin": false,
"weibo": false,
"instapaper": false,
"vk": false,
"whatsapp": false,
"all": ["facebook", "twitter", "linkedin", "weibo", "instapaper"]
},
"fontsettings": {
"theme": "white",
"family": "sans",
"size": 2
},
"edit": {
"link": "https://github.com/ehsanx/UnderstandingPropensityScore/edit/master/05-steps2.Rmd",
"text": "Edit"
},
"history": {
"link": null,
"text": null
},
"view": {
"link": null,
"text": null
},
"download": ["UnderstandingPropensityScore.pdf", "UnderstandingPropensityScore.epub"],
"search": {
"engine": "fuse",
"options": null
},
"toc": {
"collapse": "subsection"
}
});
});
</script>
<!-- dynamically load mathjax for compatibility with self-contained -->
<script>
(function () {
var script = document.createElement("script");
script.type = "text/javascript";
var src = "true";
if (src === "" || src === "true") src = "https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.9/latest.js?config=TeX-MML-AM_CHTML";
if (location.protocol !== "file:")
if (/^https?:/.test(src))
src = src.replace(/^https?:/, '');
script.src = src;
document.getElementsByTagName("head")[0].appendChild(script);
})();
</script>
</body>
</html>