David Draper

Prof. David Draper

Department of Statistics, University of California, Santa Cruz, CA 95064, USA
orcid

Research Focus & Metrics

Academic Career Length

37Years (1990-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
15 September 2016 in Journal of Business & Economic Statistics
Verified
Journal Article
Closed access
MTMatt Taddy
Matt Taddy
MGMatt Gardner
Matt Gardner
LCLiyun Chen
Liyun Chen
David DraperDavid Draper
David Draper
Randomized controlled trials play an important role in how Internet companies predict the impact of policy decisions and product changes. In these “digital experiments,” different units (people, devices, products) respond differently to the treatment. This article presents a fast and scalable Bayesian nonparametric analysis of such heterogenous treatment effects and their measurement in relation to observable covariates. New results and algorithms are provided for quantifying the uncertainty associated with treatment effect measurement via both linear projections and nonlinear regression trees (CART and random forests). For linear projections, our inference strategy leads to results that are mostly in agreement with those from the frequentist literature. We find that linear regression adjustment of treatment effect averages (i.e., post-stratification) can provide some variance reduction, but that this reduction will be vanishingly small in the low-signal and large-sample setting of digital experiments. For regression trees, we provide uncertainty quantification for the machine learning algorithms that are commonly applied in tree-fitting. We argue that practitioners should look to ensembles of trees (forests) rather than individual trees in their analysis. The ideas are applied on and illustrated through an example experiment involving 21 million unique users of EBay.com.

4 June 2018 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Verified
Journal Article
Closed access
ATAlexander Terenin
Alexander Terenin
MMMans Magnusson
Mans Magnusson
LJLeif Jonsson
Leif Jonsson
David DraperDavid Draper
David Draper
Latent Dirichlet Allocation (LDA) is a topic model widely used in natural language processing and machine learning. Most approaches to training the model rely on iterative algorithms, which makes it difficult to run LDA on big corpora that are best analyzed in parallel and distributed computational environments. Indeed, current approaches to parallel inference either don't converge to the correct posterior or require storage of large dense matrices in memory. We present a novel sampler that overcomes both problems, and we show that this sampler is faster, both empirically and theoretically, than previous Gibbs samplers for LDA. We do so by employing a novel Pólya-urn-based approximation in the sparse partially collapsed sampler for LDA. We prove that the approximation error vanishes with data size, making our algorithm asymptotically exact, a property of importance for large-scale topic models. In addition, we show, via an explicit example, that-contrary to popular belief in the topic modeling literature-partially collapsed samplers can be more efficient than fully collapsed samplers. We conclude by comparing the performance of our algorithm with that of other approaches on well-known corpora.

28 June 2004 in ACM SIGCSE Bulletin
Verified
Journal Article
Open access
BHBrian Hanks
Brian Hanks
CMCharlie McDowell
Charlie McDowell
David DraperDavid Draper
David Draper
MKMilovan Krnjajic
Milovan Krnjajic

1 May 2014 in Statistics & Probability Letters
Verified
Journal Article
Closed access
MKMilovan Krnjajić
Milovan Krnjajić
David DraperDavid Draper
David Draper

29 July 2017 in Entropy
Verified
Journal Article
Open access
PDF file
ATAlexander Terenin
Alexander Terenin
David DraperDavid Draper
David Draper
In a given problem, the Bayesian statistical paradigm requires the specification of a prior distribution that quantifies relevant information about the unknowns of main interest external to the data. In cases where little such information is available, the problem under study may possess an invariance under a transformation group that encodes a lack of information, leading to a unique prior—this idea was explored at length by E.T. Jaynes. Previous successful examples have included location-scale invariance under linear transformation, multiplicative invariance of the rate at which events in a counting process are observed, and the derivation of the Haldane prior for a Bernoulli success probability. In this paper we show that this method can be extended, by generalizing Jaynes, in two ways: (1) to yield families of approximately invariant priors; and (2) to the infinite-dimensional setting, yielding families of priors on spaces of distribution functions. Our results can be used to describe conditions under which a particular Dirichlet Process posterior arises from an optimal Bayesian analysis, in the sense that invariances in the prior and likelihood lead to one and only one posterior distribution.

1 November 2008 in International Finance Discussion Paper
Verified
Journal Article
Closed access
DBDaniel O. Beltran
Daniel O. Beltran
David DraperDavid Draper
David Draper
This paper estimates the parameters of a stylized dynamic stochastic general equilibrium model using maximum likelihood and Bayesian methods, paying special attention to the issue of weak parameter identification. Given the model and the available data, the posterior estimates of the weakly identified parameters are very sensitive to the choice of priors. We provide a set of tools to diagnose weak identification, which include surface plots of the log-likelihood as a function of two parameters, heat plots of the log-likelihood as a function of three parameters, Monte Carlo simulations using artificial data, and Bayesian estimation using three sets of priors. We find that the policy coefficients and the parameter governing the elasticity of labor supply are weakly identified by the data, and posterior predictive distributions remind us that DSGE models may make poor forecasts even when they fit the data well. Although parameter identification is model- and data specific, the lack of identification of some key structural parameters in a small-scale DSGE model such as the one we examine should raise a red flag to researchers trying to estimate — and draw valid inferences from — large-scale models featuring many more parameters.