BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:David Banks (Duke University)
DTSTART:20211004T150000Z
DTEND:20211004T155500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/1
DESCRIPTION:Title: The Statistical Challenges of Computational Advertising\nb
y David Banks (Duke University) as part of BIRS workshop: Statistical Meth
ods for Computational Advertising\n\n\nAbstract\nComputational advertising
is a relatively young field\, but it touches on almost every aspect of st
atistics. This talk frames the purpose of this workshop\, and details som
e of the ways in which computational advertising intersects with statistic
s.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/1/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Tim Hesterberg
DTSTART:20211004T160000Z
DTEND:20211004T165500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/2
DESCRIPTION:Title: Surveys and Big Data for Estimating Brand Lift\nby Tim Hes
terberg as part of BIRS workshop: Statistical Methods for Computational Ad
vertising\n\n\nAbstract\nGoogle Brand Lift Surveys estimates the effect of
display advertising using surveys. Challenges include imperfect A/B exper
iments\, response and solicitation bias\, discrepancy between intended and
actual treatment\, comparing treatment group users who took an action wit
h control users who might have acted\, and estimation for different slices
of the population. We approach these\nissues using a combination of indiv
idual-study analysis and meta-analysis across thousands of studies. This w
ork involves a combination of small and large data - survey responses and
logs data\, respectively.\nThere are a number of interesting and even surp
rising methodological twists. We use regression to handle imperfect A/B e
xperiments and response and solicitation biases\; we find regression to be
more stable than propensity methods. We use a particular form of regula
rization that combines advantages of L1 regularization (better predictions
) and L2 (smoothness). We use a variety of slicing methods\, that estimat
e either incremental or non-incremental effects of covariates like age and
gender that may be correlated. We bootstrap to obtain standard errors. I
n contrast to many regression settings\, where one may either resample obs
ervations or fix X and\nresample Y\, here only resampling observations is
appropriate.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/2/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Art Owen (Stanford University)
DTSTART:20211004T170000Z
DTEND:20211004T175500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/3
DESCRIPTION:Title: Efficiency of Tie-Breaker Designs\nby Art Owen (Stanford U
niversity) as part of BIRS workshop: Statistical Methods for Computational
Advertising\n\n\nAbstract\nA company can offer some sort of incentive or
gift to its best customers. Those gifts have a cost and so it is worth in
vestigating their causal impact. Causal impact can be measured by regress
ion discontinuity analysis. Since the incentive is under the control of t
he company they can also randomize around the cutoff in what is known as a
tie-breaker design. Perhaps the top 5% of customers get the gift along w
ith a randomly selected half of the next 10% of customers. We show that t
ie-breaker designs improve efficiency (versus regression discontinuity) fo
r linear regression models and they have advantages in local linear regres
sion as well. Surprisingly there is little to gain from using a sliding s
cale of gift probabilities instead of the levels 0%\, 50% and 100% that ap
pear in the tie-breaker design.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/3/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mamadou Yauck (Université du Québec à Montréal)
DTSTART:20211004T180000Z
DTEND:20211004T185500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/4
DESCRIPTION:Title: Computational Advertising: A Capture-Recapture Perspective
\nby Mamadou Yauck (Université du Québec à Montréal) as part of BIRS w
orkshop: Statistical Methods for Computational Advertising\n\n\nAbstract\n
This work is concerned with the analysis of marketing data on the activati
on of applications (apps) on mobile devices. Each application has a hashed
identification number that is specific to the device on which it has been
installed. This number can be registered by a platform at each activation
of the application. Activations on the same device are linked together us
ing the identification number. By focusing on activations that took place
at a business location one can create a capture-recapture data set about d
evices\, or more specifically their users\, that "visited" the business: t
he units are owners of mobile devices\, and the capture occasions are time
intervals such as days. In this talk\, we will present a new algorithm fo
r estimating the parameters of a capture-recapture model with a fairly lar
ge number of capture occasions and a simple parametric bootstrap variance
estimator.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/4/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ben Skraina (eBay)
DTSTART:20211004T200000Z
DTEND:20211004T202500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/5
DESCRIPTION:Title: Triumph and Tragedy in A/B tests: War Stories from Amazon\, eB
ay\, and Startups\nby Ben Skraina (eBay) as part of BIRS workshop: Sta
tistical Methods for Computational Advertising\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/5/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Anru Zhang (Duke University)
DTSTART:20211004T203000Z
DTEND:20211004T205500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/6
DESCRIPTION:Title: High-order Clustering with Application in Click-through Predic
tion\nby Anru Zhang (Duke University) as part of BIRS workshop: Statis
tical Methods for Computational Advertising\n\n\nAbstract\nIn e-commerce\,
predicting click-through for user-item pairs in a time-specific way plays
an important role in the online recommendation system. The click-through
data can be organized as an order-3 tensor\, where each entry is indexed b
y (users\, items\, time) and represents whether there is user-item interac
tion in a time period. The users/items often exhibit clustering structures
due to similar preferences/attributes. It is important to do high-order c
lustering\, i.e.\, to exploit such high-order clustering structures. The h
igh-order clustering problem also arises from applications in genomics and
social network studies. The non-convex and discontinuous nature of the hi
gh-order clustering problem pose significant challenges in both statistics
and computation.\n\nIn this talk\, we introduce a tensor block model and
the computationally efficient methods\, high-order Lloyd algorithm (HLloyd
)\, and high-order spectral clustering (HSC)\, for high-order clustering.
The local convergence of the proposed procedure is established under a mil
d sub-Gaussian noise assumption. In particular\, for the Gaussian tensor b
lock model\, we give a complete characterization of the statistical-comput
ational trade-off for achieving high-order exact clustering based on three
different signal-to-noise ratio regimes. We show the merits of the propos
ed procedures on the real online-click through data.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/6/
END:VEVENT
BEGIN:VEVENT
SUMMARY:S. Yaser Samadi (Southern Illinois University)
DTSTART:20211004T210000Z
DTEND:20211004T212500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/7
DESCRIPTION:Title: Dimension Reduction for Vector Autoregressive Models\nby S
. Yaser Samadi (Southern Illinois University) as part of BIRS workshop: St
atistical Methods for Computational Advertising\n\n\nAbstract\nThe classic
al vector autoregressive (VAR) models have been widely used to model multi
variate time series data\, because of their flexibility and ease of use. H
owever\, the VAR model suffers from overparameterization particularly whe
n the number of lags and number of time series get large. There are sever
al statistical methods of achieving dimension reduction of the parameter s
pace in VAR models. In this talk\, we introduce the reduced-rank VAR model
(Velu et al.\, 1986\; Reinsel and Velu\, 2013) which restricts the rank o
f the parameter matrix in one direction\, and the envelope VAR model (Wang
and Ding\, 2018) which is another solution to overcome the overparameteri
zation problem. Then\, we propose a new parsimonious VAR model by incorpor
ating the idea of envelope models into the reduced-rank VAR. We show the s
trength and efficacy of the proposed model by some simulation studies and
an economic dataset.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/7/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Patrick LeBlanc (Duke University)
DTSTART:20211005T150000Z
DTEND:20211005T155500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/8
DESCRIPTION:Title: An Overview of Recommender System Theory\nby Patrick LeBla
nc (Duke University) as part of BIRS workshop: Statistical Methods for Com
putational Advertising\n\n\nAbstract\nThis talk is a literature survey of
approaches that have been taken to various kinds of recommender systems.
I discuss both active and passive systems.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/8/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Deborshee Sen (University of Bath)
DTSTART:20211005T160000Z
DTEND:20211005T165500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/9
DESCRIPTION:Title: Cross-Domain Recommender Systems\nby Deborshee Sen (Univer
sity of Bath) as part of BIRS workshop: Statistical Methods for Computatio
nal Advertising\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/9/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Grace Yi (University of Western Ontario)
DTSTART:20211005T170000Z
DTEND:20211005T175500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/10
DESCRIPTION:Title: Unbiased Boosting Estimation for Censored Survival Data\n
by Grace Yi (University of Western Ontario) as part of BIRS workshop: Stat
istical Methods for Computational Advertising\n\n\nAbstract\nBoosting meth
ods have been broadly discussed for various settings\, especially for case
s with complete data. This talk concerns survival data which typically inv
olve censored responses. Three adjusted loss functions are proposed to add
ress the effects due to right-censored responses where no specific model i
s imposed\, and an unbiased boosting estimation method is developed. Theor
etical results\, including consistency and convergence\, are established.
Numerical studies demonstrate the promising finite sample performance of t
he proposed method.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/10/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Xuan Bi (University of Minnesota)
DTSTART:20211005T180000Z
DTEND:20211005T185500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/11
DESCRIPTION:Title: Improving Sales Forecasting Accuracy: A tensor factorization
approach with demand awareness\nby Xuan Bi (University of Minnesota) a
s part of BIRS workshop: Statistical Methods for Computational Advertising
\n\n\nAbstract\nDue to accessible big data collections from consumers\, pr
oducts\, and stores\, advanced sales forecasting capabilities have drawn g
reat attention from many companies especially in the retail business becau
se of its importance in decision making. Improvement of the forecasting ac
curacy\, even by a small percentage\, may have a substantial impact on com
panies' production and financial planning\, marketing strategies\, invento
ry controls\, supply chain management\, and eventually stock prices. Speci
fically\, our research goal is to forecast the sales of each product in ea
ch store in the near future. Motivated by tensor factorization methodologi
es for personalized context-aware recommender systems\, we propose a novel
approach called the Advanced Temporal Latent-factor Approach to Sales for
ecasting (ATLAS)\, which achieves accurate and individualized prediction f
or sales by building a single tensor-factorization model across multiple s
tores and products. Our contribution is a combination of: tensor framework
(to leverage information across stores and products)\, a new regularizati
on function (to incorporate demand dynamics)\, and extrapolation of tensor
into future time periods using state-of-the-art statistical (seasonal aut
o-regressive integrated moving-average models) and machine-learning (recur
rent neural networks) models. The advantages of ATLAS are demonstrated on
\\iv{eight datasets} collected by the Information Resource\, Inc.\, where
a total of 165 million weekly sales transactions from more than 1\,500 gro
cery stores over 15\,560 products are analyzed.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/11/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nianqiao (Phyllis) Ju (Purdue University)
DTSTART:20211005T200000Z
DTEND:20211005T205500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/12
DESCRIPTION:Title: Towards Cost-Efficient A/B Testing\nby Nianqiao (Phyllis)
Ju (Purdue University) as part of BIRS workshop: Statistical Methods for
Computational Advertising\n\n\nAbstract\nOnline A/B tests play an instrume
ntal role for Internet companies to improve products and technologies in a
data-driven manner. An online A/B test\, in its most straightforward form
\, can be treated as a static hypothesis test where traditional statistica
l tools such as p-values and power analysis might be applied to help decis
ion makers determine which variant performs better. However\, a static A/B
test presents both time cost and the opportunity cost for rapid product i
terations. While some works try to tackle these challenges\, no prior meth
od focuses on a holistic solution to both issues. We propose a unified fra
mework utilizing sequential analysis and multi-armed bandit to address tim
e cost and the opportunity cost of static online tests simultaneously. In
particular\, we present an imputed sequential Girshick test that accommoda
tes both streaming data and dynamic treatment allocation. The unobserved p
otential outcomes are treated as missing data and are imputed using empiri
cal averages. Focusing on the binomial model\, we demonstrate that the pro
posed imputed Girshick test achieves Type-I error and power control with b
oth a fixed allocation ratio and an adaptive allocation such as Thompson S
ampling through extensive experiments.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/12/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nathaniel Stevens (University of Waterloo)
DTSTART:20211005T210000Z
DTEND:20211005T215500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/13
DESCRIPTION:Title: Modern Design of Experiments for Computational Advertising\nby Nathaniel Stevens (University of Waterloo) as part of BIRS workshop:
Statistical Methods for Computational Advertising\n\n\nAbstract\nDesigned
experiments have long been regarded as the backbone of the scientific met
hod used as the gold standard for causal inference. Although DOE has tradi
tionally been applied in the realms of agriculture\, manufacturing\, pharm
aceutical development\, and the physical and social sciences\, in recent y
ears\, designed experiments have become commonplace within internet and te
chnology companies for product development/ improvement\, customer acquisi
tion/ retention\, and just about anything that impacts a business’s bott
om line. These online controlled experiments\, known colloquially as A/B t
ests\, provide an especially lucrative opportunity for modern advertisers
to understand market sentiment and consumer preferences. In this talk we p
rovide an overview of A/B testing and online controlled experiments and we
describe ways in which these experiments and this context differ from tha
t of classical experiments. Although this modern “backyard” (as Tukey
might call it) is somewhat under-appreciated in the field of industrial st
atistics\, we discuss several important and impactful research opportuniti
es that traditional industrial statisticians could and should get involved
with.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/13/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yiyun Luo (University of North Carolina Chapel Hill)
DTSTART:20211006T150000Z
DTEND:20211006T155500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/14
DESCRIPTION:Title: Distribution-Free Contextual Dynamic Pricing\nby Yiyun Lu
o (University of North Carolina Chapel Hill) as part of BIRS workshop: Sta
tistical Methods for Computational Advertising\n\n\nAbstract\n.: Contextu
al dynamic pricing aims to set personalized prices based on sequential int
eractions with customers. At each time period\, a customer who is interest
ed in purchasing a product comes to the platform. The customer's valuation
for the product is a linear function of contexts\, including product and
customer features\, plus some random market noise. The seller does not obs
erve the customer's true valuation\, but instead needs to learn the valuat
ion by leveraging contextual information and historical binary purchase fe
edbacks. Existing models typically assume full or partial knowledge of the
random noise distribution. In this paper\, we consider contextual dynamic
pricing with unknown random noise in the linear valuation model. Our dist
ribution-free pricing policy learns both the contextual function and the m
arket noise simultaneously. A key ingredient of our method is a novel pert
urbed linear bandit framework\, where a modified linear upper confidence b
ound algorithm is proposed to balance the exploration of market noise and
the exploitation of the current knowledge for better pricing. We establish
the regret upper bound and a matching lower bound of our policy in the pe
rturbed linear bandit framework and prove a sub-linear regret bound in the
considered pricing problem. Finally\, we show the superior performance of
our policy on simulations and a real-life auto-loan dataset\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/14/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Aiyou Chen
DTSTART:20211006T160000Z
DTEND:20211006T165500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/15
DESCRIPTION:Title: Robust Causal Inference for Incremental Return on Ad Spend wi
th Randomized Paired Geo Experiment\nby Aiyou Chen as part of BIRS wor
kshop: Statistical Methods for Computational Advertising\n\n\nAbstract\nEv
aluating the incremental return on ad spend (iROAS) of a prospective onlin
e marketing strategy has become progressively more important as advertiser
s increasingly seek to better understand the impact of their marketing dec
isions. Although randomized “geo experiments” are frequently employed
for this evaluation\, obtaining reliable estimates of the iROAS can be cha
llenging as oftentimes only a small number of highly heterogeneous units a
re used. In this talk\, we formulate a novel statistical framework for inf
erring the iROAS of online advertising in a randomized paired geo experime
nt design\, and we propose and develop a robust and distribution-free esti
mator “Trimmed Match” which adaptively trims poorly matched pairs. Usi
ng numerical simulations and real case studies\, we show that Trimmed Matc
h can be more efficient than some alternatives\, and we investigate the se
nsitivity of the estimator to some violations of its assumptions. This is
joint work with my colleague Tim Au at Google.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/15/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jason Poulos (Harvard)
DTSTART:20211006T170000Z
DTEND:20211006T173500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/16
DESCRIPTION:Title: Retrospective and Forward-Looking Counterfactual Imputation v
ia Matrix Completion\nby Jason Poulos (Harvard) as part of BIRS worksh
op: Statistical Methods for Computational Advertising\n\n\nAbstract\nI wil
l discuss the matrix completion method for counterfactual imputation in st
andard and retrospective panel data settings\, with applications to the so
cial sciences. This talk is partly based on joint work with Andrea Albanes
e (LISER)\, Andrea Mercatanti (Bank of Italy)\, and Fan Li (Duke).\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/16/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yi Guo (Duke University)
DTSTART:20211006T174000Z
DTEND:20211006T181500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/17
DESCRIPTION:Title: Multiparty Auctions without Common Knowledge\nby Yi Guo (
Duke University) as part of BIRS workshop: Statistical Methods for Computa
tional Advertising\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/17/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Maggie Mao
DTSTART:20211006T182000Z
DTEND:20211006T185500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/18
DESCRIPTION:Title: The Fight for Best Practices in Experimentation\nby Maggi
e Mao as part of BIRS workshop: Statistical Methods for Computational Adve
rtising\n\n\nAbstract\nOnline controlled experiments provide a scientific
approach to understand how product changes affect user behavior and site p
erformance. It is also called the A/B test\, and it is a golden standard t
o testify ideas\, quantify improvements\, and build causal relationships.
At eBay\, we have built a self-service experimentation platform to facilit
ate running experiments at scale. However\, challenges raise when democrat
izing experimentation and ensuring best practice (e.g.\, power analysis\,
multiple testing\, metrics with highly skewed distribution\, etc.). In the
talk\, I will introduce the challenges we are facing and our current copi
ng strategies.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/18/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Guy Aridor (Columbia University)
DTSTART:20211007T150000Z
DTEND:20211007T155500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/19
DESCRIPTION:Title: The Effect of Privacy Regulation on the Data Industry: Empiri
cal Evidence from GDP\nby Guy Aridor (Columbia University) as part of
BIRS workshop: Statistical Methods for Computational Advertising\n\n\nAbst
ract\nUtilizing a novel dataset from an online travel intermediary\, we st
udy the effects of EU’s General Data Protection Regulation (GDPR). The o
pt-in requirement of GDPR resulted in 12.5% drop in the intermediary-obser
ved consumers\, but the remaining consumers are trackable for a longer per
iod of time. These findings are consistent with privacy-conscious consumer
s substituting away from less efficient privacy protection (e.g\, cookie d
eletion) to explicit opt out—a process that would make opt-in consumers
more predictable. Consistent with this hypothesis\, the average value of t
he remaining consumers to advertisers has increased\, offsetting some of t
he losses from consumer opt-outs.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/19/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Fiammetta Menchetti (University of Florence)
DTSTART:20211007T160000Z
DTEND:20211007T165500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/20
DESCRIPTION:Title: ARIMA Models and Multivariate Bayesian Structural Models for
Causal Inference from Sales Data\nby Fiammetta Menchetti (University o
f Florence) as part of BIRS workshop: Statistical Methods for Computationa
l Advertising\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/20/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ernest Fokoue (University of Rochester)
DTSTART:20211007T170000Z
DTEND:20211007T175500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/21
DESCRIPTION:by Ernest Fokoue (University of Rochester) as part of BIRS wor
kshop: Statistical Methods for Computational Advertising\n\nAbstract: TBA\
n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/21/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ron Berman (The Wharton School)
DTSTART:20211007T180000Z
DTEND:20211007T185500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/22
DESCRIPTION:Title: Latent Stratification for Advertising Experiments\nby Ron
Berman (The Wharton School) as part of BIRS workshop: Statistical Methods
for Computational Advertising\n\n\nAbstract\n.: We develop a new estimat
or of the ATE for advertising incrementality experiments that improves pre
cision by estimating separate treatment effects for three latent strata --
customers who buy regardless of ad exposure\, those who buy only if expos
ed to ads and those who do not buy regardless. The overall ATE computed by
averaging the strata estimates has lower sampling variance than the widel
y-used difference-in-means ATE estimator. The variance is most reduced whe
n the three strata have substantially different ATEs and are relatively eq
ual in size. Estimating the latent stratified ATE for 5 catalog mailing ex
periments shows a reduction of 36-57% in the posterior variance of the est
imate.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/22/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Michael Braun (Southern Methodist University)
DTSTART:20211007T200000Z
DTEND:20211007T205500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/23
DESCRIPTION:Title: The A/B Test Deception: Divergent Delivery\, Ad Response Hete
rogeneity\, and Erroneous Inferences in Online Advertising Field Experimen
ts\nby Michael Braun (Southern Methodist University) as part of BIRS w
orkshop: Statistical Methods for Computational Advertising\n\n\nAbstract\n
.: Online advertising platforms provide tools to make it easy for adverti
sers to conduct randomized experiments: so-called “A/B Tests”. In a
targeted advertising environment\, true A-B tests are comparing two differ
ent mixtures of experimental subjects. We characterize how bias in the ag
gregate estimate of the difference between two ads’ lifts is driven by t
he interplay between heterogeneous responses to different ads and how plat
forms deliver ads to divergent subsets of users. We also identify conditio
ns for an undetectable “Simpson’s reversal\,” in which all unobserve
d types of users may prefer ad A over ad B\, but experimental results lead
the advertiser mistakenly infer that users prefer ad B over ad A.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/23/
END:VEVENT
BEGIN:VEVENT
SUMMARY:George Michailidis (University of Florida)
DTSTART:20211007T210000Z
DTEND:20211007T215500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/24
DESCRIPTION:Title: Regularized and Smooth Double Core Tensor Factorization for H
eterogeneous Data\nby George Michailidis (University of Florida) as pa
rt of BIRS workshop: Statistical Methods for Computational Advertising\n\n
\nAbstract\nTensor factorization based models have been extensively used i
n developing recommender systems. In this talk\, we introduce a general t
ensor model suitable for data analytic tasks for heterogeneous datasets\,
wherein there are joint low-rank structures within groups of observations\
, but also discriminative structures across different groups. To capture s
uch complex structures\, a double core tensor (DCOT) factorization model i
s introduced together with a family of smoothing loss functions. By levera
ging the proposed smoothing function\, the model accurately estimates the
model factors\, even in the presence of missing entries. A linearized ADMM
method is employed to solve regularized versions of DCOT factorizations\,
that avoid large tensor operations and large memory storage requirements.
The effectiveness of the DCOT model is illustrated on selected real-world
examples including image completion and recommender systems.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/24/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Simon Mak (Duke University)
DTSTART:20211008T150000Z
DTEND:20211008T155500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/25
DESCRIPTION:Title: TSEC: a framework for online experimentation under experiment
al constraints\nby Simon Mak (Duke University) as part of BIRS worksho
p: Statistical Methods for Computational Advertising\n\n\nAbstract\nThomps
on sampling is a popular algorithm for solving multi-armed bandit problems
. In many applications\, however\, the number of choices (or arms) can be
large\, and the data needed to make adaptive decisions require expensive e
xperimentation. One is then faced with the constraint of experimenting on
only a small subset of arms within each time period\, which poses a proble
m for traditional Thompson sampling. To address this\, we propose a new Th
ompson Sampling under Experimental Constraints (TSEC) method\, which makes
use of a Bayesian interaction model to model reward correlations between
different arms. This fitted model is then integrated within Thompson sampl
ing\, to jointly identify a good subset of arms for experimentation and to
allocate resources over these arms. We demonstrate the effectiveness of T
SEC in two applications with arm budget constraints: the first on website
optimization\, and the second for portfolio optimization.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/25/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sammy Natour
DTSTART:20211008T160000Z
DTEND:20211008T165500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/26
DESCRIPTION:by Sammy Natour as part of BIRS workshop: Statistical Methods
for Computational Advertising\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/26/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Edoardo Airoldi
DTSTART:20211008T170000Z
DTEND:20211008T175500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/27
DESCRIPTION:Title: Estimating Peer-Influence Effects Under Homophily: Randomized
Treatments and Insights\nby Edoardo Airoldi as part of BIRS workshop:
Statistical Methods for Computational Advertising\n\n\nAbstract\n: Classi
cal approaches to causal inference largely rely on the assumption of lack
of interference\, according to which the outcome of an individual does not
depend on the treatment assigned to others\, as well as on many other sim
plifying assumptions\, including the absence of strategic behavior. In man
y applications\, however\, such as evaluating the effectiveness of health-
related interventions that leverage social structure\, assessing the impac
t of product innovations and ad campaigns on social media platforms\, or e
xperimentation at scale in large IT organizations\, several common simplif
ying assumptions are simply untenable. Moreover\, being able to quantify a
spects of complications\, such as the causal effect of interference itself
\, are often inferential targets of interest\, rather than nuisances. In t
his talk\, we will formalize issues that arise in estimating causal effect
s when interference can be attributed to a network among the units of anal
ysis\, within the potential outcomes framework. We will introduce and disc
uss several strategies for experimental design in this context centered ar
ound a useful role for statistical models. In particular\, we wish for cer
tain finite-sample properties of the estimates to hold even if the model c
atastrophically fails\, while we would like to gain efficiency if certain
aspects of the model are correct. We will then contrast design-based\, mod
el-based and model-assisted approaches to experimental design from a decis
ion theoretic perspective.\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/27/
END:VEVENT
BEGIN:VEVENT
SUMMARY:David Banks
DTSTART:20211008T180000Z
DTEND:20211008T185500Z
DTSTAMP:20260404T060943Z
UID:BIRS-21w5508/28
DESCRIPTION:Title: Closing Remarks\nby David Banks as part of BIRS workshop:
Statistical Methods for Computational Advertising\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/BIRS-21w5508/28/
END:VEVENT
END:VCALENDAR