A Comparison of Two Linear Discriminant Analysis Methods That Use Block Monotone Missing Training Data

We revisit a comparison of two discriminant analysis procedures, namely the linear combination classifier of Chung and Han (2000) and the maximum likelihood estimation substitution classifier for the problem of classifying unlabeled multivariate normal observations with equal covariance matrices into one of two classes. Both classes have matching block monotone missing training data. Here, we demonstrate that for intra-class covariance structures with at least small correlation among the variables with missing data and the variables without block missing data, the maximum likelihood estimation substitution classifier outperforms the Chung and Han (2000) classifier regardless of the percent of missing observations. Specifically, we examine the differences in the estimated expected error rates for these classifiers using a Monte Carlo simulation, and we compare the two classifiers using two real data sets with monotone missing data via parametric bootstrap simulations. Our results contradict the conclusions of Chung and Han (2000) that their linear combination classifier is superior to the MLE classifier for block monotone missing multivariate normal data.

A General Hermitian Nonnegative-Definite Solution to the Matrix Equation <i>AXB</i>= <i>C</i>

We derive necessary and sufficient conditions for the existence of a Hermitian nonnegative-definite solution to the matrix equation AXB = C. Moreover, we derive a representation of a general Hermitian nonnegative-definite solution. We then apply our solution to two examples, including a comparison of our solution to a proposed solution by Zhang in [1] using an example problem given from [1]. Our solution demonstrates that the proposed general solution from Zhang in [1] is incorrect. We also give a second example in which we derive the general covariance structure so that two matrix quadratic forms are independent.

Linear Dimension Reduction for Multiple Heteroscedastic Multivariate Normal Populations

For the case where all multivariate normal parameters are known, we derive a new linear dimension reduction (LDR) method to determine a low-dimensional subspace that preserves or nearly preserves the original feature-space separation of the individual populations and the Bayes probability of misclassification. We also give necessary and sufficient conditions which provide the smallest reduced dimension that essentially retains the Bayes probability of misclassification from the original full-dimensional space in the reduced space. Moreover, our new LDR procedure requires no computationally expensive optimization procedure. Finally, for the case where parameters are unknown, we devise a LDR method based on our new theorem and compare our LDR method with three competing LDR methods using Monte Carlo simulations and a parametric bootstrap based on real data.

Bayesian Approach to Ranking and Selection for a Binary Measurement System

Binary measurement systems that classify parts as either pass or fail are widely used. Inspectors or inspection systems are often subject to error. The error rates are unlikely to be identical across inspectors. We propose a random effects Bayesian approach to model the error probabilities and overall conforming rate. We also introduce a feature-subset selection procedure to determine the best inspector in terms of overall classification accuracy. We provide simulation studies that demonstrate the viability of our proposed estimation ranking and subset-selection methods and apply the methods to a real data set.

Dimension Reduction for Detecting a Difference in Two High-Dimensional Mean Vectors

We consider the efficacy of a proposed linear-dimension-reduction method to potentially increase the powers of five hypothesis tests for the difference of two high-dimensional multivariate-normal population-mean vectors with the assumption of homoscedastic covariance matrices. We use Monte Carlo simulations to contrast the empirical powers of the five high-dimensional tests by using both the original data and dimension-reduced data. From the Monte Carlo simulations, we conclude that a test by Thulin [1], when performed with post-dimension-reduced data, yielded the best omnibus power for detecting a difference between two high-dimensional population-mean vectors. We also illustrate the utility of our dimension-reduction method real data consisting of genetic sequences of two groups of patients with Crohn’s disease and ulcerative colitis.