Journal of Data Science

With the growing scale of big datasets, fitting novel statistical models on larger-than-memory datasets becomes correspondingly challenging. This document outlines the development and use of an API for large scale modelling, with a demonstration given by the proof of concept platform largescaler, developed specifically for the development of statistical models for big datasets.

This paper provides an overview of how to use “big data” for social science research (with an emphasis on economics and finance). We investigate the performance and ease of use of different Spark applications running on a distributed file system to enable the handling and analysis of data sets which were previously not usable due to their size. More specifically, we explain how to use Spark to (i) explore big data sets which exceed retail grade computers memory size and (ii) run typical statistical/econometric tasks including cross sectional, panel data and time series regression models which are prohibitively expensive to evaluate on stand-alone machines. By bridging the gap between the abstract concept of Spark and ready-to-use examples which can easily be altered to suite the researchers need, we provide economists and social scientists more generally with the theory and practice to handle the ever growing datasets available. The ease of reproducing the examples in this paper makes this guide a useful reference for researchers with a limited background in data handling and distributed computing.

Penalized regression provides an automated approach to preform simultaneous variable selection and parameter estimation and is a popular method to analyze high-dimensional data. Since the conception of the LASSO in the mid-to-late 1990s, extensive research has been done to improve penalized regression. The LASSO, and several of its variations, performs penalization symmetrically around zero. Thus, variables with the same magnitude are shrunk the same regardless of the direction of effect. To the best of our knowledge, sign-based shrinkage, preferential shrinkage based on the sign of the coefficients, has yet to be explored under the LASSO framework. We propose a generalization to the LASSO, asymmetric LASSO, that performs sign-based shrinkage. Our method is motivated by placing an asymmetric Laplace prior on the regression coefficients, rather than a symmetric Laplace prior. This corresponds to an asymmetric ${\ell _{1}}$ penalty under the penalized regression framework. In doing so, preferential shrinkage can be performed through an auxiliary tuning parameter that controls the degree of asymmetry. Our numerical studies indicate that the asymmetric LASSO performs better than the LASSO when effect sizes are sign skewed. Furthermore, in the presence of positively-skewed effects, the asymmetric LASSO is comparable to the non-negative LASSO without the need to place an a priori constraint on the effect estimates and outperforms the non-negative LASSO when negative effects are also present in the model. A real data example using the breast cancer gene expression data from The Cancer Genome Atlas is also provided, where the asymmetric LASSO identifies two potentially novel gene expressions that are associated with BRCA1 with a minor improvement in prediction performance over the LASSO and non-negative LASSO.

Multi-classification is commonly encountered in data science practice, and it has broad applications in many areas such as biology, medicine, and engineering. Variable selection in multiclass problems is much more challenging than in binary classification or regression problems. In addition to estimating multiple discriminant functions for separating different classes, we need to decide which variables are important for each individual discriminant function as well as for the whole set of functions. In this paper, we address the multi-classification variable selection problem by proposing a new form of penalty, supSCAD, which first groups all the coefficients of the same variable associated with all the discriminant functions altogether and then imposes the SCAD penalty on the supnorm of each group. We apply the new penalty to both soft and hard classification and develop two new procedures: the supSCAD multinomial logistic regression and the supSCAD multi-category support vector machine. Our theoretical results show that, with a proper choice of the tuning parameter, the supSCAD multinomial logistic regression can identify the underlying sparse model consistently and enjoys oracle properties even when the dimension of predictors goes to infinity. Based on the local linear and quadratic approximation to the non-concave SCAD and nonlinear multinomial log-likelihood function, we show that the new procedures can be implemented efficiently by solving a series of linear or quadratic programming problems. Performance of the new methods is illustrated by simulation studies and real data analysis of the Small Round Blue Cell Tumors and the Semeion Handwritten Digit data sets.

Clustering is an essential technique for discovering patterns in data. Many clustering algorithms have been developed to tackle the ever increasing quantity and complexity of data, yet algorithms that can cluster data with mixed variables (continuous and categorical) remain limited despite the abundance of mixed-type data. Of the existing clustering methods for mixed data types, some posit unverifiable distributional assumptions or rest on unbalanced contributions of different variable types. To address these issues, we propose a two-step hybrid density- and partition-based (HyDaP) algorithm to detect clusters after variable selection. The first step involves both density-based and partition-based algorithms to identify the data structure formed by continuous variables and determine important variables (both continuous and categorical) for clustering. The second step involves a partition-based algorithm together with our proposed novel dissimilarity measure to obtain clustering results. Simulations across various scenarios were conducted to compare the HyDaP algorithm with other commonly used methods. Our HyDaP algorithm was applied to identify sepsis phenotypes and yielded important results.