Thursday, Aug 8: 10:30 AM - 12:20 PM
5199
Contributed Papers
Oregon Convention Center
Room: CC-G129
Main Sponsor
Section on Nonparametric Statistics
Presentations
In this article, we present a novel nonparametric class of life distributions, using the concept of starshaped functions, to address challenges encountered in reliability and life testing. Our focus is on scenarios where the increasing or decreasing Mean Residual Life (IMRL or DMRL) aging classes prove inadequate in capturing the underlying aging process. Motivated by practical challenges in applied reliability, we introduce a test procedure designed to assess whether the overall remaining life distribution exhibits a decreasing Mean Residual Life (MRL) property, specifically concerning the exponential distribution.The proposed test accommodates right-censored and randomly censored data as well. We conduct simulation studies to evaluate the empirical power of the test. This work aims to provide a valuable tool for modeling aging processes in situations where the IMRL concept falls short, but overall lifelengths demonstrate improvement with age. We believe that our approach contributes to the broader understanding of reliability analysis and life testing, addressing gaps left by traditional aging classes.
Keywords
Starshaped function
Renewal Process
U-statistics
Equilibrium distribution
Mean residual life distribution
Stochastic process
Popular regression methods for clustered survival data become inadequate when functional forms of covariates and interactions are unknown. Spatially correlated random cluster effects and unknown cluster-level covariates further complicate the issue. In this context, we propose a nonparametric approach, SBART (soft Bayesian additive regression trees) within the Bayesian ensemble learning paradigm (Basak et al., 2022). Our method addresses challenges like numerous clusters, variable cluster sizes, and data information for proper statistical augmentation of the unobserved covariate sourced from a data registry different from the survival study. We illustrate the practical implementation of our method and its advantages over existing methods by assessing the impacts of intervention in some county level and patient-level covariates to eliminate existing racial disparity in breast cancer survival in different Florida counties. The sources of clustered survival data with patient-level covariates and the data information for an unobserved county-level covariate are different. We also compare our method with existing analysis methods to demonstrate our advantages through simulation studies.
Keywords
Soft Bayesian Additive Regression Trees (SBART)
Data Augmentation
Disparity
Markov Chain Monte Carlo (MCMC)
Metropolis Hastings (MH) Algorithm.
Unobserved covariate
Controversial posts from social media users during major national or international events reflect real-time reactions to topics discussed on the platform. Capturing shifts in public morale from social media posts can effectively detect public perceptions of particular events and help guide public responses. Assessing changes in sentiment of social media posts from users who are opinion leaders can help evaluate the impact of information disseminated in the cycle. We collected user tweets from Brandwatch, transformed the observed sentiment series into transition pairs, and utilized sparse multivariate functional data analysis to model the transitions among sentiments from Twitter posts on conversations related to Israel-Hams conflict, reflecting the immediate effects of information diffusion. Rather than solely considering users' sentiments over time, this work focuses on switches in sentiment that represent an instant reaction toward the information consumed from social media platforms. We aim to provide interpretable clustering memberships representing reactions to information from different opinions to a specific crisis event over time, and identify potential individuals who are amenable to the information diffusion. This approach can be generalized to cluster users from any social media platform for any event, and it is implemented in the \texttt{clustersc()} function from the \texttt{R} package \texttt{catfda}.
Keywords
shift
crisis event
markov chain
sparse multivariate functional data
Social Media
Abstracts
Multispectral camera images have been playing a crucial role in animal vision research. Reconstructing animal vision from static images typically involves camera calibration, spectral reflectance estimation, and linear transformation of camera responses to animal quantum catches - a procedure that lacks transferability across scenes as camera recalibration is required for changes in illumination settings. In this study, we propose a novel yet simple framework for studying animal vision using multispectral consumer cameras. This framework requires additional estimation of the spectral illumination using a basis representation, with the advantage of needing only a one-time camera calibration. Unlike typical regression problems involving basis functions, only camera readings are observed, which are the inner products of spectral illumination, sensitivity, and reflectance. To address this, the coefficients are estimated using transformed basis functions, and Bayesian interval estimation is applied. This framework enables the reconstruction of animal vision with moving objects and variable lighting, paving the ways for generating animal vision videos in natural habitats.
Keywords
multispectral imaging
function estimation
constrained estimation
P-splines
Bayesian inference
The Gini distance correlation (GDC) is a recently proposed dependence measure to assess the relationship between a categorical variable and a numerical variable. GDC has been found to possess more attractive properties than existing measures of dependence. In this study, we develop the jackknife empirical likelihood (JEL) approach for the GDC. We then build confidence intervals for the correlation without estimating the asymptotic variance. In addition, we explore adjusted and weighted JEL methods to enhance the standard JEL's performance. Simulation studies demonstrate that our approaches are competitive with existing methods in terms of coverage accuracy and the shortness of confidence intervals. The proposed methods are illustrated using a real-data example.
Keywords
Gini distance correlation
Confidence interval
Jackknife empirical likelihood
Co-Author
Yongli Sang, University of Louisiana at Lafayette
First Author
Sameera Hewage, University of Louisiana at Lafayette
Presenting Author
Sameera Hewage, University of Louisiana at Lafayette
In many prediction problems, the predictive model affects the sampling distribution. For example, job applicants may doctor their resumes to get them past resume screening systems. Such distribution shifts are especially common in social computing, but methods for learning such distribution shifts from data are scarce. Drawing inspiration from a microeconomic model that characterizes agents in labor markets, we introduce a novel approach for learning the distribution shift. This approach assumes that the agents in the population are rational and adapt to the predictive model by choosing from a finite set of actions. We demonstrate, both empirically and theoretically, that learning the distribution shift and plugging it into an existing method for performative prediction, which necessitates knowledge of the distribution map, is (statistically) more efficient than zeroth-order methods that avoid learning the distribution map.
Keywords
Strategic classification
Performative prediction
Distribution shift
Cost estimation
Tests for uniformity on the circle based on spacings, such as the Rao's spacing test, assume that the data are measured on a continuous scale. In practice however, recorded observations are often rounded values, ignoring which can result in incorrect inference, especially if the degree of rounding is quite severe or if the sample size is large. In this article, we propose a simple modification to the rounded data that allows us to continue to use the Rao's spacing test and its available critical values, without affecting the probability of Type I error. In support of the theoretical justification, we provide simulation studies to demonstrate how our proposed method works and evaluate its performance.
Keywords
circular data
spacings
rounding
test for uniformity
Rao's spacing test