Browsing by Subject "Sampling"

Television (TV) is the predominant advertising medium, and recent technological advances such as digital video recorders (DVRs) and set-top boxes (STBs) have the potential to transform this industry by enabling household-specific advertising. Since exposure to TV represents a substantial share of consumer time and attention, this potential to micro-target communications represents an enormous opportunity for the TV advertising market.

This paper outlines an approach to facilitate the micro-targeting of TV advertising. We employ a unique dataset, integrating TV program and advertisement viewing at the household level with purchase data, to address the question of how advertisers can achieve better advertising targeting in the digital context. Based on this dataset, we first develop a model of household TV viewing behavior. The viewing model comprises three integrated components: TV show sampling and watching, TV show recording, and advertising viewing. All three components are motivated by the theoretical concept of flow utility, that is, the moment-by-moment enjoyment a household derives from different activities: watching a TV show, watching a TV advertisement, and other non-TV activities. This model has decent out-of-sample prediction power on show choices and time spent on each selected show. We then link household advertising exposure with purchase. Finally, the viewing model and identified advertising-sales relationship are utilized to conduct counterfactual policy experiments on advertising targeting. We consider several household-level targeting scenarios by manipulating: 1) whether the advertising purchase is made in advance; and 2) whether the objective function is to minimize costs for a given set of exposures or to maximize revenues from advertising. Results indicate micro-targeting can lower advertising costs and raise incremental revenue.

The key contributions of this paper are as follows. Theoretically, we develop an integrated model on TV show viewing, TV advertising viewing, purchasing and advertising targeting. Methodologically, we propose a new modeling framework on media consumption by explicitly accounting for the role of uncertainty, and propose targeting strategies leveraging household-level data. Substantively, we offer policy recommendations to advertisers on micro-targeting which can be of great potential.

In this thesis, we study the convergence of diffusion models and related flow-based methods, which are highly successful approaches for learning a probability distribution from data and generating further samples. For diffusion models, we established the first convergence result applying to data distributions satisfying the log-sobolev inequality without suffering the curse of dimensionality. Our analysis gives theoretical grounding to the observation that an annealed procedure is required in practice to generate good samples, as our proof depends essentially on using annealing to obtain a warm start at each step. Moreover, we show that a predictor-corrector algorithm gives better convergence than using either portion alone. Then we generalized the results to any distribution with bounded 2nd moment, relying only on a $L^2$-accurate score estimates, with polynomial dependence on all parameters and no reliance on smoothness or functional inequalities. We also provide a theoretical guarantee of generating data distribution by a progressive flow model, the so-called JKO flow model, which implements the Jordan-Kinderleherer-Otto (JKO) scheme in a normalizing flow network. Leveraging the exponential convergence of the proximal gradient descent (GD) in Wasserstein space, we provethe Kullback-Leibler (KL) guarantee of data generation by a JKO flow model where the assumption on data density is merely a finite second moment

South Africa has a substantial HIV epidemic as well as a rising methamphetamine use problem, particularly in Cape Town. Respondent driven sampling (RDS) may be a useful tool for engaging vulnerable and hard-to-reach populations in HIV research, although its effectiveness has not yet been examined among South African methamphetamine users. The aim of the current study was to describe the effectiveness of RDS as a method for engaging methamphetamine users in Cape Town into a HIV behavioral research study. RDS procedures were used to screen 374 potential participants from a peri-urban township in Cape Town. Measures of homophily, equilibrium and RDS-1 estimators were computed for key demographic and social variables.

Beginning with 8 seeds, 345 methamphetamine users were enrolled over a 6 month period, with a coupon return rate of 67%. The sample included 197 men and 148 women who were ethnically diverse (73% Coloured, 27% Black African) and had a mean age of 28.8 years (SD=7.2). Social networks were adequate (mean network size >5) and mainly comprised of close social ties. Equilibrium on race was reached after 11 waves of recruitment, and after ≤3 waves for all other variables of interest. There was little to moderate preference for either in- or out-group recruiting in all subgroups.

Results suggest that RDS is an effective method for engaging methamphetamine users into HIV prevention research in South Africa. RDS may be a useful strategy for seeking high risk methamphetamine users for HIV testing and linkage to HIV care in this and other low resource settings. We also discuss future directions for RDS studies.

Many biological processes are effectively modeled as networks, but a frequent assumption is that these networks do not change during data collection. However, that assumption does not hold for many phenomena, such as neural growth during learning or changes in genetic regulation during cell differentiation. Approaches are needed that explicitly model networks as they change in time and that characterize the nature of those changes.

In this work, we develop a new class of graphical models in which the conditional dependence structure of the underlying data-generation process is permitted to change over time. We first present the model, explain how to derive it from Bayesian networks, and develop an efficient MCMC sampling algorithm that easily generalizes under varying levels of uncertainty about the data generation process. We then characterize the nature of evolving networks in several biological datasets.

We initially focus on learning how neural information flow networks change in songbirds with implanted electrodes. We characterize how they change in response to different sound stimuli and during the process of habituation. We continue to explore the neurobiology of songbirds by identifying changes in neural information flow in another habituation experiment using fMRI data. Finally, we briefly examine evolving genetic regulatory networks involved in Drosophila muscle differentiation during development.

We conclude by suggesting new experimental directions and statistical extensions to the model for predicting novel neural flow results.

The study of stratified sampling is of interest in systems which canbe solved accurately on small scales, or which depend heavily on rare transitions of particles from one subspace to another. We present a new form of stratified MCMC algorithm built with non-reversible stochastic dynamics in mind. The method has potential usefulness in that many systems of interest are non-reversible, and can also benefit from stratification at the same time. It may also be useful for sampling on complex manifolds, and hence manifold learning. Our method is a generalization of previous stratified or nested sampling schemes which extend QSD sampling schemes. It can also be viewed as a generalization of the exact milestoning method previously studied by D. Aristoff. The primary advantages of our new results over such previous studies are generalization to non-reversible processes, expressions for the convergence rate in terms of the process's behavior within each stratum and large scale behavior between strata, and less restrictive assumptions for convergence. We show that the algorithm has a unique fixed point which corresponds to the invariant measure of the process without stratification. We will show how the speeds of two versions of the new algorithm, one with an extra eigenvalue problem step and one without, relate to the mixing rate of a discrete process on the strata, and the mixing probability of the process being sampled within each stratum. The eigenvalue problem version also relates to local and global perturbation results of discrete Markov chains, such as those given by J. Weare. Finally, we will propose a way to relate the accuracy of finite approximations of a process using our stratified scheme to its expected exit times from each stratum and its approximation of the true process's generator, by means of a Poisson equation argument.