We propose a nonparametric method to perform functional principal components analysis for the case of sparse longitudinal data. The aims at irregularly spaced data, where number repeated measurements available per subject is small. In contrast, classical data requires large regularly subject. assume that are located randomly with random repetitions each and determined by an underlying smooth (subject-specific) trajectory plus measurement errors. Basic elements our approach parsimonious...

With the advance of modern technology, more and data are being recorded continuously during a time interval or intermittently at several discrete points. These both examples functional data, which has become commonly encountered type data. Functional analysis (FDA) encompasses statistical methodology for such Broadly interpreted, FDA deals with theory that in form functions. This paper provides an overview FDA, starting simple notions as mean covariance functions, then covering some core...

We propose nonparametric methods for functional linear regression which are designed sparse longitudinal data, where both the predictor and response functions of a covariate such as time. Predictor processes have smooth random trajectories, data consist small number noisy repeated measurements made at irregular times sample subjects. In studies, per subject is often may be modeled discrete and, accordingly, only finite asymptotically nonincreasing available each or experimental unit....

The use of principal component methods to analyze functional data is appropriate in a wide range different settings. In studies “functional analysis,” it has often been assumed that sample random functions observed precisely, the continuum and without noise. While this traditional setting for analysis, context longitudinal analysis function typically represents patient, or subject, who at only small number randomly distributed points, with nonnegligible measurement error. Nevertheless,...

Nonparametric estimation of mean and covariance functions is important in functional data analysis. We investigate the performance local linear smoothers for both with a general weighing scheme, which includes two commonly used schemes, equal weight per observation (OBS), subject (SUBJ), as special cases. provide comprehensive analysis their asymptotic properties on unified platform all types sampling plan, be it dense, sparse or neither. Three are investigated this paper: normality, $L^{2}$...

Although variation in mortality is considered by virtually all vector-borne disease specialists to be one of the most important determinants an arthropod's capacity transmit pathogens, operational assumption often that insect vector independent age. Acceptance non-senescence leads erroneous conclusion mosquito age unimportant, results misleading predictions regarding reductions after control, and represses study other aspects biology change with We brought large-scale laboratory life table...

With the advance of modern technology, more and data are being recorded continuously during a time interval or intermittently at several discrete points. They both examples "functional data", which have become prevailing type data. Functional Data Analysis (FDA) encompasses statistical methodology for such Broadly interpreted, FDA deals with analysis theory that in form functions. This paper provides an overview FDA, starting simple notions as mean covariance functions, then covering some...

We present a probabilistic framework for studying adversarial attacks on discrete data. Based this framework, we derive perturbation-based method, Greedy Attack, and scalable learning-based Gumbel that illustrate various tradeoffs in the design of attacks. demonstrate effectiveness these methods using both quantitative metrics human evaluation state-of-the-art models text classification, including word-based CNN, character-based CNN an LSTM. As as example our results, show accuracy...

Deep neural networks obtain state-of-the-art performance on a series of tasks. However, they are easily fooled by adding small adversarial perturbation to the input. The is often imperceptible humans image data. We observe significant difference in feature attributions between adversarially crafted examples and original examples. Based this observation, we introduce new framework detect through thresholding scale estimate attribution scores. Furthermore, extend our method include multi-layer...

The life history of medflies is characterized by two physiological modes with different demographic schedules fertility and survival: a waiting mode in which both mortality reproduction are low reproductive very at the onset egg laying but accelerates as eggs laid. Medflies stay when they fed only sugar. When protein, scarce resource wild, switch to mode. that from survive longer than kept either exclusively. An understanding shift occurs between may yield information about fundamental...

Journal Article Joint modelling of accelerated failure time and longitudinal data Get access Yi-Kuan Tseng, Tseng Department Statistics, University California, Davis, California 95616, U.S.A. yktseng@wald.ucdavis.edu, fushing@wald.ucdavis.edu, wang@wald.ucdavis.edu Search for other works by this author on: Oxford Academic Google Scholar Fushing Hsieh, Hsieh Jane-Ling Wang Biometrika, Volume 92, Issue 3, September 2005, Pages 587–603, https://doi.org/10.1093/biomet/92.3.587 Published: 01 2005

Without parametric assumptions, high-dimensional regression analysis is already complex. This made even harder when data are subject to censoring. In this article, we seek ways of reducing the dimensionality regressor before applying nonparametric smoothing techniques. If censoring time independent lifetime, then method sliced inverse can be applied directly. Otherwise, modification needed adjust for bias. A key identity leading bias correction derived and root-$n$ consistency modified...

Classical multivariate principal component analysis has been extended to functional data and termed (FPCA). Most existing FPCA approaches do not accommodate covariate information, it is the goal of this paper develop two methods that do. In first approach, both mean covariance functions depend on Z time scale t while in second approach only function depends Z. Both new additional measurement errors sampled at regular grids as well sparse longitudinal irregular grids. The fully adjust adapts...

In many situations, data are recorded over a period of time and may be regarded as realizations stochastic process. this paper, robust estimators for the principal components considered by adapting projection pursuit approach to functional setting. Our combines projection-pursuit with different smoothing methods. Consistency shown under mild assumptions. The performance classical procedures compared in simulation study contamination schemes.

Ongoing variability in neural signaling is an intrinsic property of the brain. Often this considered to be noise and ignored. However, alternative view that fundamental perception cognition may particularly important decision-making. Here, we show a momentary measure occipital alpha-band power (8-13 Hz) predicts choices about where human participants will focus spatial attention on trial-by-trial basis. This finding provides evidence for mechanistic account decision-making by demonstrating...

While deep learning approaches to survival data have demonstrated empirical success in applications, most of these methods are difficult interpret and mathematical understanding them is lacking. This paper studies the partially linear Cox model, where nonlinear component model implemented using a neural network. The proposed approach flexible able circumvent curse dimensionality, yet it facilitates interpretability effects treatment covariates on survival. We establish asymptotic theories...

Abstract Quantifying the association between components of multivariate random curves is general interest and a ubiquitous basic problem that can be addressed with functional data analysis. An important application assessing connectivity based on magnetic resonance imaging (fMRI), where one aims to determine similarity fMRI time courses are recorded anatomically separated brain regions. In literature, static temporal Pearson correlation has been prevailing measure for connectivity. However,...

Summary We propose a class of semiparametric functional regression models to describe the influence vector-valued covariates on sample response curves. Each observed curve is viewed as realization random process, composed an overall mean function and components. The finite dimensional components eigenfunction expansion through single-index that include unknown smooth link variance functions. parametric are estimated via quasi-score estimating equations with functions being nonparametrically....

Left truncation and right censoring arise frequently in practice for life data. This paper is concerned with the estimation of hazard rate function such Two types nonparametric estimators based on kernel smoothing methods are considered. The first one obtained by convolving a cumulative estimator. second form ratio two statistics. Local properties including consistency, asymptotic normality mean squared error expressions presented both estimators. These facilitate locally adaptive bandwidth...

A new single-index model that reflects the time-dynamic effects of single index is proposed for longitudinal and functional response data, possibly measured with errors, both time-invariant covariates. With appropriate initial estimates parametric index, estimator shown to be $\sqrt{n}$-consistent asymptotically normally distributed. We also address nonparametric estimation regression functions provide optimal convergence rates. One advantage approach same bandwidth used estimate mean...

Mount Everest is an extreme environment for humans. Nevertheless, hundreds of mountaineers attempt to summit each year. In a previous study we analyzed interview data all climbers (2,211) making their first on during 1990-2005. Probabilities summiting were similar men and women, declined progressively about 40 older, but elevated with experience climbing in Nepal. dying also increased 60 older (especially the few that had summited), independent experience. Since 2005, many more (3,620) have...