ABSTRACT: The mode of a distribution provides an important summary of data and is often estimated on the basis of some non‐parametric kernel density estimator. This article develops a new data analysis tool called modal linear regression in order to explore high‐dimensional data. Modal linear regression models the conditional mode of a response Y given a set of predictors x as a linear function of x. Modal linear regression differs from standard linear regression in that standard linear regression models the conditional mean (as opposed to mode) of Y as a linear function of x. We propose an expectation–maximization algorithm in order to estimate the regression coefficients of modal linear regression. We also provide asymptotic properties for the proposed estimator without the symmetric assumption of the error density. Our empirical studies with simulated data and real data demonstrate that the proposed modal regression gives shorter predictive intervals than mean linear regression, median linear regression and MM‐estimators.

High-dimensional feature selection arises in many areas of modern science. For example, in genomic research we want to find the genes that can be used to separate tissues of different classes (e.g. cancer and normal) from tens of thousands of genes that are active (expressed) in certain tissue cells. To this end, we wish to fit regression and classification models with a large number of features (also called variables, predictors). In the past decade, penalized likelihood methods for fitting regression models based on hyper-LASSO penalization have received increasing attention in the literature. However, fully Bayesian methods that use Markov chain Monte Carlo (MCMC) are still in lack of development in the literature. In this paper we introduce an MCMC (fully Bayesian) method for learning severely multi-modal posteriors of logistic regression models based on hyper-LASSO priors (non-convex penalties). Our MCMC algorithm uses Hamiltonian Monte Carlo in a restricted Gibbs sampling framework; we call our method Bayesian logistic regression with hyper-LASSO (BLRHL) priors. We have used simulation studies and real data analysis to demonstrate the superior performance of hyper-LASSO priors, and to investigate the issues of choosing heaviness and scale of hyper-LASSO priors.

The problem of selecting the most useful features from a great many (eg, thousands) of candidates arises in many areas of modern sciences. An interesting problem from genomic research is that, from thousands of genes that are active (expressed) in certain tissue cells, we want to find the genes that can be used to separate tissues of different classes (eg. cancer and normal). In this paper, we report our empirical experiences of using Bayesian logistic regression based on heavy-tailed priors with moderately small degree freedom (such as 1) and very small scale, and using Hamiltonian Monte Carlo to do computation. We discuss the advantages and limitations of this method, and illustrate the difficulties that remain unsolved. The method is applied to a real microarray data set related to prostate cancer. The method identifies only 3 non-redundant genes out of 6033 candidates but achieves better leave-one-out cross-validated prediction accuracy than many other methods.

Feature selection is demanded in many modern scientific research problems that use high-dimensional data. A typical example is to find the most useful genes that are related to a certain disease (eg, cancer) from high-dimensional gene expressions. The expressions of genes have grouping structures, for example, a group of co-regulated genes that have similar biological functions tend to have similar expressions. Many statistical methods have been proposed to take the grouping structure into consideration in feature selection, including group LASSO, supervised group LASSO, and regression on group representatives. In this paper, we propose a fully Bayesian Robit regression method with heavy-tailed (sparsity) priors (shortened by FBRHT) for selecting features with grouping structure. The main features of FBRHT include that it discards more aggressively unrelated features than LASSO, and it can make feature selection within groups automatically without a pre-specified grouping structure. In this paper, we use simulated and real datasets to demonstrate that the predictive power of the sparse feature subsets selected by FBRHT are comparable with other much larger feature subsets selected by LASSO, group LASSO, supervised group LASSO, penalized logistic regression and random forest, and that the succinct feature subsets selected by FBRHT have significantly better predictive power than the feature subsets of the same size taken from the top features selected by the aforementioned methods.