-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
executable file
·400 lines (357 loc) · 22.3 KB
/
index.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
<!DOCTYPE html>
<html lang="en">
<head>
<!-- Google Analytics -->
<!-- Global Site Tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-106900339-1"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments)};
gtag('js', new Date());
gtag('config', 'UA-106900339-1');
</script>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1">
<!-- The above 3 meta tags *must* come first in the head; any other head content must come *after* these tags -->
<meta name="description" content="">
<meta name="author" content="">
<link rel="icon" href="../../favicon.ico">
<title>Alexander D'Amour</title>
<!-- Bootstrap core CSS -->
<link href="css/bootstrap.min.css" rel="stylesheet">
<!-- Bootstrap theme -->
<link href="css/bootstrap-theme.min.css" rel="stylesheet">
<!-- IE10 viewport hack for Surface/desktop Windows 8 bug -->
<link href="assets/css/ie10-viewport-bug-workaround.css" rel="stylesheet">
<!-- Custom styles for this template -->
<link href="theme.css" rel="stylesheet">
<!-- Just for debugging purposes. Don't actually copy these 2 lines! -->
<!--[if lt IE 9]><script src="../../assets/js/ie8-responsive-file-warning.js"></script><![endif]-->
<script src="js/ie-emulation-modes-warning.js"></script>
<!-- HTML5 shim and Respond.js for IE8 support of HTML5 elements and media queries -->
<!--[if lt IE 9]>
<script src="https://oss.maxcdn.com/html5shiv/3.7.2/html5shiv.min.js"></script>
<script src="https://oss.maxcdn.com/respond/1.4.2/respond.min.js"></script>
<![endif]-->
</head>
<body role="document">
<div class="container" role="main">
<div class="row" style="padding:20px">
<div class="hidden-xs col-sm-3 col-md-2" id="sidebar" role="navigation",
style = "margin:20px 0 0 0">
<ul class="nav nav-pills nav-stacked">
<li><a href="blog/public/index.html">Blog</a></li>
<li><a href="#papers">Papers</a></li>
<li><a href="#talks">Talks and Media</a></li>
<li><a href="#teaching">Teaching</a></li>
<li><a href="#consulting">Consulting</a></li>
</ul>
</div>
<div class="col-xs-10 col-sm-8 col-md-8">
<div class="row">
<img src="headshot2.jpg" class="pull-left" style="margin:20px
20px 20px 0; height:200px; width:200px" />
<h1>Alexander D'Amour</h1>
<p class="lead">
Staff Research Scientist, Google DeepMind<br>Cambridge, MA<br>
alexdamour@google.com<br>
</p>
<p>
<ul class="list-inline">
<li/><a href="content/damour_cv.pdf">Curriculum Vitae</a>
<li/><a href="https://scholar.google.com/citations?user=okP0uukAAAAJ">Google Scholar</a>
<li/><a href="https://twitter.com/alexdamour">Twitter</a>
<li/><a href="blog/public/index.html">Blog</a>
</p>
</div><!-- row -->
<hr>
<div class="row">
<p>I am a Staff Research Scientist at Google DeepMind in Cambridge, MA. Formerly, I was a Neyman Visiting Assistant Professor in the Department of Statistics at UC Berkeley. I did my PhD in the Department of Statistics at Harvard University, where I was advised by <a href="http://www.people.fas.harvard.edu/~airoldi/">Edoardo Airoldi</a>.
<!--I was a member of the <a href="http://applied.stat.harvard.edu">Harvard Laboratory for Applied Statistical Methodology & Data Science</a>.-->
</p>
<p>
I work primarily on problems in causal inference, distribution shift, domain generalization, and fairness for building reliable, trustworthy Machine Learning/AI systems. More generally, I am interested in problems where simple prediction is not enough.
</p>
<p></p>
I am interested in all kinds of applications. In research and in consulting, I've worked on problems in sports, healthcare, education, social network analysis, marketing, finance, microfinance, and entertainment.
</p>
<!--<p></p>
<p>I am an active member of the <a href="http://xyresearch.com" data-url="http://xyresearch.com">XY Research</a> group, which conducts research in sports statistics with a focus on player-tracking data.</p>
<p></p>-->
</div>
<!--
<div class="row"><h2 class="text-muted">Details</h2>
<hr>
<ul class="list-inline">
<li><a href="content/research_statement.pdf">Research Statement</a>
<li><a href="content/teaching_statement.pdf">Teaching Statement</a>
</ul>
<p><strong>Dissertation. </strong>My dissertation research was about the statistical analysis of social network data, particularly the logical difficulties that arise from the sparse scaling behavior of social networks. This work spans the full stack arguments and methods that are employed in a scientific investigation of social networks, from the logical role that misspecified models play in an investigation, to new modeling and inference methodologies for drawing predictive and causal inferences from the network data.</p>
<p><strong>Application areas. </strong>Through my academic and consulting work, I have completed projects in a wide range of fields, including document disambiguation, text analysis, epidemiology, education technology, digital marketing, customer modeling in e-commerce, and credit access in developing economies.</p>
<p><strong>Background.</strong> I also hold AB and SM degrees in Applied Mathematics, also from Harvard University.</p>
</div>
-->
<div class="row">
<h2 class="text-muted"><a name="papers"></a>Papers</h2>
<p>
<i>Note: This section is no longer maintained. Please see <a href="https://scholar.google.com/citations?user=okP0uukAAAAJ">Google Scholar</a> for up-to-date publications.</i>
</p>
<hr>
<p><span class="lead">To Appear</span></p>
<p>
<strong>On Multi-Cause Causal Inference with Unobserved Confounding: Counterexamples, Impossibility, and Alternatives</strong><br>
<span class="text-muted">Contrary to some recent claims, evaluating many causes (e.g., treatments) simultaneously cannot eliminate unobserved confounding in observational causal inference. This paper demonstrates this point by counterexample</span><br>
<strong>Alexander D'Amour</strong><br>
<em>To appear in Proceedings of AISTATS 2019</em>
<ul class="list-inline">
<li><a href="https://arxiv.org/abs/1902.10286">arXiv Preprint</a>
</ul>
</p>
<p>
<strong>Flexible sensitivity analysis for observational studies without observable implications</strong><br>
<span class="text-muted">In causal inference, sensitivity analysis is meant to probe unidentifiable assumptions that are necessary for causal identification, but may sensitivity analysis methods inadvertently impose restrictions that have observable implications. We propose a flexible, interpretable framework that does not have this problem and is compatible with modern observed data modeling techniques.</span><br>
Alexander Franks, <strong>Alexander D'Amour</strong>, and Avi Feller<br>
<em>To appear in the Journal of the American Statistical Association</em>
<ul class="list-inline">
<li><a href="https://arxiv.org/abs/1809.00399">arXiv Preprint</a>
</ul>
</p>
<p><span class="lead">Under Review</span></p>
<p>
<strong>Overlap in Observational Studies with High-Dimensional Covariates</strong><br>
<span class="text-muted">In high dimensions, overlap is a stronger assumption than most people realize. This paper presents some implications.</span><br>
<strong>Alexander D'Amour</strong>, Peng Ding, Avi Feller, Lihua Lei, and Jasjeet Sekhon<br>
<ul class="list-inline">
<!--<li><a href="content/effective_ancillarity/effective_ancillarity.html">Colloquium Slides</a>-->
<li><a href="https://arxiv.org/abs/1711.02582">arXiv Preprint</a>
</ul>
</p>
<p><span class="lead">Published</span></p>
<p>
<strong>Reducing Reparameterization Gradient Variance</strong><br>
<span class="text-muted">Control variate technique for reducing the variance of stochastic gradients used in Monte Carlo variational inference.</span><br>
Andrew C. Miller, Nicholas J. Foti, <strong>Alexander D'Amour</strong>, and Ryan P. Adams<br>
<em>Advances in Neural Information Processing Systems (NIPS), 2017 </em>
<ul class="list-inline">
<li><a href="https://arxiv.org/abs/1705.07880">arXiv Preprint</a>
</ul>
</p>
<p>
<strong>Meta-Analytics: Tools for Understanding the Statistical Properties of Sports Metrics</strong><br>
<span class="text-muted">Introduces an ensemble of r-squared-style statistics to quantify the reliability and uniqueness of sports metrics.</span><br>
Alexander Franks, <strong>Alexander D'Amour</strong>, Daniel Cervone, and Luke Bornn<br>
<em>Journal of Quantitative Analysis in Sports</em>
<ul class="list-inline">
<li><a href="https://arxiv.org/abs/1609.09830">arXiv Preprint</a>
<li><a href="https://www.degruyter.com/view/j/jqas.2016.12.issue-4/jqas-2016-0098/jqas-2016-0098.xml">Journal Paper</a>
</ul>
</p>
<p>
<strong>A Multiresolution Stochastic Process Model for Predicting Basketball Possession Outcomes</strong><br>
<span class="text-muted">Methodology for computing Expected Possession Value, an instantaneous expected point value for a basketball possession.</span><br>
Daniel Cervone, <strong>Alexander D'Amour</strong>, Luke Bornn, and Kirk Goldsberry<br>
<em>Journal of the American Statistical Association</em>
<ul class="list-inline">
<li><a href="http://arxiv.org/abs/1408.0777">arXiv Preprint</a>
<li><a href="http://www.tandfonline.com/doi/suppl/10.1080/01621459.2016.1141685">Journal Paper</a>
<li><a href="http://www.sloansportsconference.com/?p=13017">SSAC Presentation and Paper</a>
<li><a href="content/EPV_poster.pdf">ISBA 2014 Poster<a>
<li><a href="http://grantland.com/features/expected-value-possession-nba-analytics/">Grantland Feature</a>
</ul>
</p>
<p>
<strong>Disambiguation and Co-authorship Networks of the U.S. Patent Inventor Database</strong><br>
<span class="text-muted">A supervised learning approach to adding unique inventor identifiers to the US patent database.</span><br>
G. Li, R. Lai, <strong>Alexander D'Amour</strong>, D. Doolin, Y. Sun, V. Torvik, A. Yu, and L. Fleming<br>
<em>Research Policy</em>, 2014.
<ul class="list-inline">
<li><a href="http://www.sciencedirect.com/science/article/pii/S0048733314000225">Journal Paper</a>
</ul>
</p>
<p>
<strong>Estimating Rates of Carriage Acquisition and Clearance and Competitive Ability for Pneumococcal Serotypes in Kenya With a Markov Transition Model</strong><br>
<span class="text-muted">Markov model approach to estimating epideiological properties of <em>Pneumococcal</em> serotypes using periodic testing data from Kenyan schoolchildren.</span><br>
M. Lipsitch, O. Abdullani, <strong>Alexander D'Amour</strong>, W. Xie, D. Weinberger, E. Tchetgen, and J. Scott<br>
<em>Epidemiology</em>, 2012.
<ul class="list-inline">
<li><a href="http://www.ncbi.nlm.nih.gov/pubmed/22441543">Journal Paper</a>
</ul>
</p>
<p>
<strong>Improving Major League Park Factor Estimates</strong><br>
<span class="text-muted">An ANOVA approach to estimating park factors in Major League Baseball. Written in conjunction with the Harvard Sports Analysis Collective.</span><br>
R. Acharya, A. Ahmed, <strong>Alexander D'Amour</strong>, H. Lu, C. Morris, B. Oglevee, A. Peterson, and R. Swift<br>
<ul class="list-inline">
<li><a href="http://www.degruyter.com/view/j/jqas.2008.4.2/jqas.2008.4.2.1108/jqas.2008.4.2.1108.xml">Journal Paper</a>
</ul>
</p>
<p><span class="lead">Dissertation</span></p>
<p>
<strong>The Effective Estimand</strong><br>
<span class="text-muted">A framework for characterizing the scientific usefulness of an estimator derived from a misspecified model.
Generalizes the work on networks to general modeling tasks.</span><br>
<strong>Alexander D'Amour</strong> and Edoardo Airoldi<br>
<ul class="list-inline">
<li class="text-danger">In preparation.
<!--<li><a href="content/effective_ancillarity/effective_ancillarity.html">Colloquium Slides</a>-->
<li><a href="content/effective_estimand_prearxiv.pdf">Working Draft</a>
</ul>
</p>
<p>
<strong>Misspecification, Sparsity, and Superpopulation Inference for Sparse Social Networks</strong><br>
<span class="text-muted">Theoretical characterization of how the sparse scaling of social networks undermines superpopulation investigations when the sparsity is not modeled exactly.
Proposes sparsity-invariant modeling and inference methodology.</span><br>
<strong>Alexander D'Amour</strong> and Edoardo Airoldi<br>
<ul class="list-inline">
<li class="text-danger">In preparation.
<li><a href="content/sparsity_misspecification_prearxiv.pdf">Working Draft</a>
<li><a href="content/damour_hsph_2-22-2016.pdf">Slides</a>
</ul>
</p>
<p>
<strong>Causal Inference with Sparse Social-Interaction-Valued Outcomes</strong><br>
<span class="text-muted">Extension of sparsity-invariant methodology for network data to causal settings.</span><br>
<strong>Alexander D'Amour</strong> and Edoardo Airoldi<br>
<ul class="list-inline">
<li class="text-danger">In preparation.
<li><a href="content/itxval_causal_prearxiv.pdf">Working Draft</a>
</ul>
</p>
<!--<p>
<strong>Measuring the Causal Effect of the Michigan Anti-trust Reform Act of 1986 on Inventor Collaboration Dynamics in Michigan</strong><br>
<span class="text-muted">Sparsity-invariant methodology applied in a causal policy evaluation using the US patent coauthorship network.</span>
<strong>Alexander D'Amour</strong>, Edoardo Airoldi, and Lee Fleming<br>
<ul class="list-inline">
<li class="text-danger">In preparation.
</ul>
</p>-->
</div>
<div class="row"><a name="talks"></a><h2 class="text-muted">Talks, Posters, Other Media</h2>
<hr>
<p><span class="lead">Talks</span></p>
<p>
<strong>Overlap in High Dimensions</strong><br>
<span class="text-muted">Surprisingly strong implications of the overlap assumption that is usually invoked in high-dimensional causal inference. Upshot: in high dimensions, the overlap assumption approaches a balance assumption.
</span><br/>
Invited talk at the <em>Berkeley Division of Biostatistics Seminar</em>, October 2017 at UC Berkeley.<br/>
Invited talk at the <em>Atlantic Causal Inference Conference</em>, May 2017 at UNC Chapel Hill.
<ul class="list-inline">
<li><a href="https://docs.google.com/presentation/d/1UBQ5C3N7VdsKSLo5hQywjFQdxVdKUlDwybMfMN3Ftlc/pub?start=false&loop=false&delayms=3000&slide=id.p">Slides</a>
</ul>
</p>
<p>
<strong>Advances in Basketball Analytics Using Player-Tracking Data</strong><br>
<span class="text-muted">High-level overview of new quantiative methods for understanding basketball, implemented by XY Research group using player-tracking data from the NBA. </span><br>
Invited talk at <em><a href="https://cdar.berkeley.edu">Consortium for Data Analytics and Risk</a></em>, October 2017 at UC Berkeley.<br/>
Invited talk at <em>Boston ML</em> meetup, July 2016 in Boston, MA.
<ul class="list-inline">
<li><a href="https://www.youtube.com/watch?v=4kqTBO5KDr4">Video</a>
</ul>
</p>
<p>
<strong>Prediction is Not Enough: Designing decision-support statistics for causal inference and attribution</strong><br>
<span class="text-muted">Exploration of Statistical applications where the objective requires more than the ability to predict future replications of the observe data stream.</span><br>
Invited talk at <em>Clarify Health Solutions</em> in San Francisco, CA.<br/>
Invited talk at <em>Lumos Labs</em> in San Francisco, CA.<br>
<ul class="list-inline">
<li><a href="https://docs.google.com/presentation/d/1qxsMZ2vkOcBsbnDt_GDbm_POVLzNNtAM2ZdfflaVm20/pub?start=false&loop=false&delayms=3000">Slides</a>
</ul>
</p>
<p>
<strong>A Design-Based Perspective on Variable Selection</strong><br>
<span class="text-muted">An approach to variable selection that treats it as the design choice -- namely choosing which conditional distribution to model. Some preliminary thoughts on optimal data-splitting.</span><br>
Talk given in the Harvard Statistics Department's Research in Statistics student colloquium.<br>
<ul class="list-inline">
<li><a href="content/variable_selection/PQ_VS_slides.html">Colloquium Slides</a>
</ul>
</p>
<p><span class="lead">Posters</span></p>
<p>
<strong>Extrapolation Parameterizations for Assessing Sensitivity to Unmeasured Confounding</strong><br>
<span class="text-muted">Proposes the extrapolation factorization for sensitivity analysis in causal inference, which explicitly separates identified and unidentified parts of the potential outcomes model.</span><br>
<ul class="list-inline">
<li><a href="content/ACICExtrapolationPoster.pdf">Poster</a>
</ul>
</p>
<p>
<strong>Move or Die: How Ball Movement Creates Open Shots in the NBA</strong><br>
<span class="text-muted">Uses summaries of a Markov model for basketball possessions to show that ball movement is effective only inasmuch as it introduced <em>unpredictability</em> into an NBA offense.</span><br>
<span class="text-danger">Winner: Best Poster, 2015 Sloan Sports Analytics Conference.</span><br>
<ul class="list-inline">
<li><a href="content/MoveOrDie.pptx">Poster</a>
</ul>
</p>
<p><span class="lead">Popular Media</span></p>
<p>
<strong>Bayesian Statistician</strong><br>
<em>You're the Expert</em> (radio show)<br>
<ul class="list-inline">
<li><a href="http://www.podcasts.com/youre-the-expert/episode/bayesian">Podcast</a>
</ul>
</p>
<p>
<strong>Behind Databall: A Discussion on the Methodology of Expected Possession Value</strong><br>
<em>Grantland</em>
<ul class="list-inline">
<li><a href="http://grantland.com/the-triangle/behind-databall-a-discussion-on-the-methodology-of-expected-points-value/">Article</a>
</ul>
</p>
</div>
<div class="row"><a name="teaching"></a><h2 class="text-muted">Teaching</h2>
<hr>
<p><span class='lead'>Classes</span></p>
<p>
At Berkeley, I have taught the following courses:
</p>
<ul>
<li><strong>Statistics 153</strong>: Timeseries Analysis (Spring 2017, Fall 2017)
<li><strong>Statistics 298, 278B</strong>: Causal Inference Reading Group (Fall 2016, Spring 2017, Fall 2017, Spring 2018)
<li><strong>Statistics 88</strong>: Probability and Mathematical Statistics for Data Science (Fall 2016)
</ul>
<p>
At Harvard, I was a teaching fellow for the following courses:
</p>
<ul>
<li><strong>Statistics 220</strong>: Bayesian Data Analysis (Fall 2011, Fall 2012)
<li><strong>Statistics 221</strong>: Statistical Computation and Visualization (Spring 2013)
<li><strong>Statistics 225</strong>: Spatial Statistics (Spring 2014)
<li><strong>Statistics 121/Computer Science 109</strong>: Data Science (Fall 2013, Fall 2014)
<li><strong>Statistics 107</strong>: Financial Statistics (Spring 2012)
</ul>
<p><span class='lead'>Awards</span></p>
<ul>
<li><a href="http://www.stat.harvard.edu/Academics/PickardAwards/">2014 David Pickard Memorial Teaching Fellow</a>.
<li>Four-time awardee of the <a href="http://bokcenter.harvard.edu/certificates-excellence">Certificate for Distinction in Teaching</a>.
</ul>
</div>
<div class="row"><a name="consulting"></a><h2 class="text-muted">Consulting</h2>
<hr>
In a previous life, I fielded many applied statistical problems from industry in a Data Science consulting practice. I was a founding partner of Damyata, LLC, a consultancy that I founded with two tech industry veterans. Our mission was to establish best practices in Data Science by delivering state-of-the art data-driven systems to our clients. A core part of our mission was to foster academic-industry research partnerships.</p>
<p>Former consulting clients include
<ul>
<li><a href="https://www.clarifyhealth.com/">Clarify Health Solutions</a>
<li><a href="https://www.nba.com/sixers">Philadelphia 76ers</a>
<li><a href="https://www.blueapron.com/">Blue Apron</a>
<li><a href="https://www.legendary.com/">Legendary Pictures</a>
<li><a href="https://www.firstaccessmarket.com/">First Access</a>
<li><a href="https://www.demandsignals.com/">Demand Signals</a>
<li><a href="https://www.knewton.com/">Knewton, Inc</a>
<li><a href="https://www.hsph.harvard.edu/">Harvard T.H. Chan School of Public Health</a>
</ul>
</div>
</div> <!-- col -->
</div> <!-- row -->
<hr>
</div> <!-- /container -->
<!-- Bootstrap core JavaScript
================================================== -->
<!-- Placed at the end of the document so the pages load faster -->
<script src="https://ajax.googleapis.com/ajax/libs/jquery/1.11.3/jquery.min.js"></script>
<script>window.jQuery || document.write('<script src="js/vendor/jquery.min.js"><\/script>')</script>
<script src="js/bootstrap.min.js"></script>
<script src="js/docs.min.js"></script>
<!-- IE10 viewport hack for Surface/desktop Windows 8 bug -->
<script src="js/ie10-viewport-bug-workaround.js"></script>
</body>
</html>