Some Uses for Distribution-Fitting Software in Teaching Statistics

Statistics courses now make extensive use of menu-driven, interactive computer software. This article presents some insight as to how a new class of PC-based statistical software, called “distribution-fitting” software, can be used in teaching various courses in statistics.     

Partial Partial Likelihood

The maximum likelihood and maximum partial likelihood approaches to the proportional hazards model are unified. The purpose is to give a general approach to the analysis of the proportional hazards model, whether the baseline distribution is absolutely continuous, discrete, or a mixture. The advantage is that heavily tied data will be analyzed with a discrete time model, while data with no ties is analyzed with ordinary Cox regression. Data sets in between are treated by a compromise between the discrete time model and Efron's approach to tied data in survival analysis, and the transitions between modes are automatic. A simulation study is conducted comparing the proposed approach to standard methods of handling ties. A recent suggestion, that revives Breslow's approach to tied data, is finally discussed.     

Copula Density Estimation by Total Variation Penalized Likelihood

Copulas are full measures of dependence among random variables. They are increasingly popular among academics and practitioners in financial econometrics for modeling comovements between markets, risk factors, and other relevant variables. A copula's hidden dependence structure that couples a joint distribution with its marginals makes a parametric copula non-trivial. An approach to bivariate copula density estimation is introduced that is based on a penalized likelihood with a total variation penalty term. Adaptive choice of the amount of regularization is based on approximate Bayesian Information Criterion (BIC) type scores. Performance are evaluated through the Monte Carlo simulation.     

Moment-Based Approximations of Discrete Probability Distributions Using Rational Functions

Provost et al. (2009 Provost, S. B., Jiang, M. and Ha, H.-T. 2009. Moment-based approximations of probability mass functions with applications involving order statistics. Communications in Statistics—Theory and Methods, 38(12): 1969–1981. [Taylor & Francis Online], [Web of Science ®] , [Google Scholar]) introduced an approximation to discrete probability mass functions using a base density function and a polynomial function multiplier. In this article, we examine a new estimator constructed by using a rational function multiplier. The new approximation requires fewer moments to calculate, and generally performs better. Several example cases will be presented.     

Statistical Analysis for the Inverse Trinomial Distribution

This article considers parameter estimation, goodness of fit, likelihood ratio and score tests, and model selection by Akaike information criterion for the inverse trinomial (IT) distribution, a classical one-dimensional random walk distribution. The IT distribution has a cubic variance function of the mean and is a generalization of the negative binomial distribution. Basic distributional properties and expressions for the probability mass function, recurrence formula, moments, and score functions are also presented.     

Asymptotic Distribution Theory for Nonparametric Entropy Measures of Serial Dependence

Entropy is a classical statistical concept with appealing properties. Establishing asymptotic distribution theory for smoothed nonparametric entropy measures of dependence has so far proved challenging. In this paper, we develop an asymptotic theory for a class of kernel‐based smoothed nonparametric entropy measures of serial dependence in a time‐series context. We use this theory to derive the limiting distribution of Granger and Lin's (1994) normalized entropy measure of serial dependence, which was previously not available in the literature. We also apply our theory to construct a new entropy‐based test for serial dependence, providing an alternative to Robinson's (1991) approach. To obtain accurate inferences, we propose and justify a consistent smoothed bootstrap procedure. The naive bootstrap is not consistent for our test. Our test is useful in, for example, testing the random walk hypothesis, evaluating density forecasts, and identifying important lags of a time series. It is asymptotically locally more powerful than Robinson's (1991) test, as is confirmed in our simulation. An application to the daily S&P 500 stock price index illustrates our approach.