Another Cautionary Note about R 2: Its Use in Weighted Least-Squares Regression Analysis

A recent article in this journal presented a variety of expressions for the coefficient of determination (R ²) and demonstrated that these expressions were generally not equivalent. The article discussed potential pitfalls in interpreting the R ² statistic in ordinary least-squares regression analysis. The current article extends this discussion to the case in which regression models are fit by weighted least squares and points out an additional pitfall that awaits the unwary data analyst. We show that unthinking reliance on the R ² statistic can lead to an overly optimistic interpretation of the proportion of variance accounted for in the regression. We propose a modification of the estimator and demonstrate its utility by example.     

Things that make us different: analysis of deviance with time-use data

The constrained, non-normal nature of time-use data poses a challenge to ordinary analysis of variance. This paper investigates a computationally simple variance decomposition technique suitable for those data. As a by-product of the analysis, a measure of fit for systems of time-demand equations is proposed that possesses several useful properties.     

The Empirical Likelihood Goodness-of-Fit Test for a Regression Model with Randomly Censored Data

The regression model with randomly censored data has been intensively investigated. In this article, we consider a goodness-of-fit test for this model. Empirical likelihood (EL) tests are constructed. The asymptotic distributions of the test statistic under null hypothesis and the local alternative hypothesis are given. Simulations are carried out to illustrate the methodology.     

Diagnosis of Multivariate Control Chart Signal Based on Dummy Variable Regression Technique

Abstract

It is common to monitor several correlated quality characteristics using the Hotelling's T ² statistic. However, T ² confounds the location shift with scale shift and consequently it is often difficult to determine the factors responsible for out of control signal in terms of the process mean vector and/or process covariance matrix. In this paper, we propose a diagnostic procedure called ‘D-technique’ to detect the nature of shift. For this purpose, two sets of regression equations, each consisting of regression of a variable on the remaining variables, are used to characterize the ‘structure’ of the ‘in control’ process and that of ‘current’ process. To determine the sources responsible for an out of control state, it is shown that it is enough to compare these two structures using the dummy variable multiple regression equation. The proposed method is operationally simpler and computationally advantageous over existing diagnostic tools. The technique is illustrated with various examples.     

On confidence regions for the mean of a multivariate time series

We consider the problem of setting up a confidence region for the mean of amultivariate timeseries ont he basis of a part-realisation of that series.A procedure for setting up a confidence interval for the mean of a univariate time series Is implicitin Jones(1976).We present an analogous procedure for setting up a confidence region for the mean of a multivariatet ime series.This procedure is base donastatistic which is an analogue of Hotelling'sT'.Our results are applied to a comparison of climate means obtained from experiments with a General Circulation Model of the earth's atmosphere.     

A General Class of Distributions: Properties and Applications

ABSTRACT

In this article, we derive a general class of distributions and establish its relationship to χ² distribution. The proposed class includes normal, inverse Gaussian, lognormal, gamma, Rayleigh, and Maxwell distributions. Various statistical properties of the class are discussed. Some applications of the class are given.     

On the distribution of hotelling's t2 and multiple correlation r2when sampling from a mixture of two normals

In this paper the non-null distribution of Hotelling's T² and the null distribution of multiple correlation R² are derived when the sample is taken from a mixture of two p-component multivariate normal distributions with mean vectors μ₁ and μ₂ respectively and common covariance matrix ∑, ∑. In a special case the non-null distribution of R² is a l s o given, while the general noncentral distribution is given i n Awan (1981). These results have been used to study the robustness of T² and R² tests by Srivastava and Awan (1982), and Awan and Srivastava (1982) respectively.     

The Exact General Formulas for the Moments of a Ridge Regression Estimator when the Regression Error Terms Follow a Multivariate t Distribution

Huang (1999 Huang , J. C. ( 1999 ). Improving the estimation precision for a selected parameter in multiple regression analysis: an algebraic approach . Econ. Lett. 62 : 261 – 264 .[Crossref], [Web of Science ®] , [Google Scholar]) proposed a feasible ridge regression (FRR) estimator to estimate a specific regression coefficient. Assuming that the error terms follow a normal distribution, Huang (1999 Huang , J. C. ( 1999 ). Improving the estimation precision for a selected parameter in multiple regression analysis: an algebraic approach . Econ. Lett. 62 : 261 – 264 .[Crossref], [Web of Science ®] , [Google Scholar]) examined the small sample properties of the FRR estimator. In this article, assuming that the error terms follow a multivariate t distribution, we derive an exact general formula for the moments of the FRR estimator to estimate a specific regression coefficient. Using the exact general formula, we obtain exact formulas for the bias, mean squared error (MSE), skewness, and kurtosis of the FRR estimator. Since these formulas are very complex, we compare the bias, MSE, skewness, and kurtosis of the FRR estimator with those of ordinary least square (OLS) estimator by numerical evaluations. Our numerical results show that the range of MSE dominance of the FRR estimator over the OLS estimator is widen under a fat tail distributional assumption.     

Comparison of Goodness-of-Fit Measures in Probit Regression Model

This article examines several goodness-of-fit measures in the binary probit regression model. Existing pseudo-R ² measures are reviewed, two modified and one new pseudo-R ² measure are proposed. For the probit regression model, empirical comparisons are made for different goodness-of-fit measures with the squared sample correlation coefficient of the observed response and the predicted probabilities. As an illustration, the goodness-of-fit measures are applied to a “paid labor force” data set.     

On a test of independence for contingency tables

Using the concept of distributional distance, a test statistic is proposed FOR the hypothesis of independence in multidimensional contingency tables. A Monte Carlo Study is done to empirically compare the power of the proposed test to the Pearson x² and the likelihood ratio test- Further, the nonnull distribution under various spike alternatives is tabulated     

Goodness of fit for thei ordered categories discrete uniform distribution

Goodness of fit for thei ordered categories discrete uniform distribution can be carried out using Pearson's X² _pstatistic and its components. Applications of this technique are considered and comparisons made with recently suggested empirical uniform distribution     

A Simultaneous Confidence Band for Dense Longitudinal Regression

We present a method of using local linear smoothing to construct simultaneous confidence bands for the mean function of densely spaced functional data. Our approach works well under mild conditions. In addition, the local linear estimator and its accompanying confidence band enjoy semiparametric efficiency in the sense that they are asymptotically equivalent to the counterparts obtained from the random trajectories entirely observed without errors. We illustrate the performance of the proposed confidence band through a simulation study. Furthermore, an application in food science is presented.     

A Coefficient of Determination for Generalized Linear Models

The coefficient of determination, a.k.a. R², is well-defined in linear regression models, and measures the proportion of variation in the dependent variable explained by the predictors included in the model. To extend it for generalized linear models, we use the variance function to define the total variation of the dependent variable, as well as the remaining variation of the dependent variable after modeling the predictive effects of the independent variables. Unlike other definitions that demand complete specification of the likelihood function, our definition of R² only needs to know the mean and variance functions, so applicable to more general quasi-models. It is consistent with the classical measure of uncertainty using variance, and reduces to the classical definition of the coefficient of determination when linear regression models are considered.     

Goodness-of-Fit Statistics for General Linear Regression Equations in the Presence of Replicated Responses

Goodness-of-fit statistics for general multiple-linear-regression equations are reviewed for the case of replicated responses. A modification of the coefficient of determination is recommended. This statistic has 1.0 as its achievable upper bound and has the coefficient of determination as a special case. It indicates more effectively how close a general-linear-regression equation is relative to the best possible one and is particularly useful when the purpose is to ascertain whether higher-order terms of a given set of explanatory variables are required. Other goodness-of-fit statistics that take into account the variation within replicated responses are reviewed. An illustration example is presented.     

Asymptotic expaxsioxs for the joint distribution of cirrelated hotellings t2 statlstics under normality

Let T² _i=z′_iS^?1z_i, i==,…k be correlated Hotelling's T² statistics under normality. where z=(z′_i,…,z′_k)′ and nS are independently distributed as N_kp((O,ρ?∑) and Wishart distribution W_p(∑, n), respectively. The purpose of this paper is to study the distribution function F(x₁,…,x_k) of (T² _i,…,T² _k) when n is large. First we derive an asymptotic expansion of the characteristic function of (T² _i,…,T² _k) up to the order n^?2. Next we give asymptotic expansions for (T² _i,…,T² _k) in two cases (i)ρ=I_k and (ii) k=2 by inverting the expanded characteristic function up to the orders n^?2 and n^?1, respectively. Our results can be applied to the distribution function of max (T² _i,…,T² _k) as a special case.     

The coefficient of determination and its adjusted version in linear regression models

This article presents a comparative study of the efficiency properties of the coefficient of determination and its adjusted version in linear regression models when disturbances are not necessarily normal.     

Empirical Comparison of Nonparametric Regression Estimates on Real Data

The performance of nine different nonparametric regression estimates is empirically compared on ten different real datasets. The number of data points in the real datasets varies between 7, 900 and 18, 000, where each real dataset contains between 5 and 20 variables. The nonparametric regression estimates include kernel, partitioning, nearest neighbor, additive spline, neural network, penalized smoothing splines, local linear kernel, regression trees, and random forests estimates. The main result is a table containing the empirical L₂ risks of all nine nonparametric regression estimates on the evaluation part of the different datasets. The neural networks and random forests are the two estimates performing best. The datasets are publicly available, so that any new regression estimate can be easily compared with all nine estimates considered in this article by just applying it to the publicly available data and by computing its empirical L₂ risks on the evaluation part of the datasets.     

On the behaviour of some transforms of the sample correlation coefficient in samples from the bivariate t and the bivariate X2distribution

The present paper studies the normality of five transformations suggested in the literature to normalize the sample correlation coefficient. The parent populations are the bivariate t and the bivariate X ²The results in the previous work of Subrahmaniam and Gajjar are exploited to assess their performance. The density estimation procedure of Tarter and Kronmal is used to provide empiric support to the asymptotic results