LOCAL INFLUENCE IN NONLINEAR REPRODUCTIVE DISPERSION MODELS

Nonlinear reproductive dispersion models (NRDM, Jorgensen 1997) include a wider range of distributions and nonlinear models such as the possibility of correlated errors and nonlinear hypotheses dropping the exponential family assumption. Based on the generalized Cook distance and the conformal normal curvature of Poon & Poon (1999), local influence of minor perturbations on the data set is investigated for NRDM. Two examples are used to illustrate our results.     

Detection of influential observations in nonlinear reproductive dispersion models

A robust case-deletion diagnostic which could avoid masking is presented for nonlinear reproductive dispersion models (NRDM) (Jorgensen, 1997) that include generalized linear models and exponential family nonlinear models as special cases. Based on the second-order approximation of log-likelihood displacement and Poon & Poon's (1999) conformal normal curvature, a measure of local influence ranging from 0 to 1 is constructed. An example illustrates application of the techniques.     

Local Conditional Influence

Through an investigation of normal curvature functions for influence graphs of a family of perturbed models, we develop the concept of local conditional influence. This concept can be used to study masking and boosting effects in local influence. We identify the situation under which the influence graph of the unperturbed model contains all the information on these effects. The linear regression model is used for illustration and it is shown that the concept developed is consistent with Lawrance's (1995) approach of conditional influence in Cook's distance.     

A Note on Local Influence Based on Normal Curvature

Object functions other than the likelihood displacement, such as a parameter estimate or a test statistic, can also be used in local influence analysis. The normal curvatures of these object functions have been studied in situations where the slopes were non-zero. In these situations, we show that the normal curvature is not scale invariant and thus ambiguous conclusions will be drawn. Comments on the application of the general normal curvature formula are presented.     

Influence diagnostics in a vector autoregressive model

In this paper, we use a likelihood approach and the local influence method introduced by Cook [Assessment of local influence (with discussion). J Roy Statist Soc Ser B. 1986;48:133–149] to study a vector autoregressive (VAR) model. We present the maximum likelihood estimators and the information matrix. We establish the normal curvature and slope diagnostics for the VAR model under several perturbation schemes and use the Monte Carlo method to obtain benchmark values for determining the influence of directional diagnostics and possible influential observations. An empirical study using the VAR model to fit real data of monthly returns of IBM and S&P500 index illustrates the effectiveness of our proposed diagnostics.     

Sensitivity analysis of partially linear models with response missing at random

This article investigates case-deletion influence analysis via Cook’s distance and local influence analysis via conformal normal curvature for partially linear models with response missing at random. Local influence approach is developed to assess the sensitivity of parameter and nonparametric estimators to various perturbations such as case-weight, response variable, explanatory variable, and parameter perturbations on the basis of semiparametric estimating equations, which are constructed using the inverse probability weighted approach, rather than likelihood function. Residual and generalized leverage are also defined. Simulation studies and a dataset taken from the AIDS Clinical Trials are used to illustrate the proposed methods.     

Diagnostics for Autocorrelated Regression Models

This paper discusses the local influence approach to the linear regression model with AR(1) errors. Diagnostics for the autocorrelation models and for the autocorrelation coefficient only are proposed and developed respectively, when simultaneous perturbations of the response vector are allowed. Furthermore, the direction of maximum curvature of local influence analysis is shown to be exactly the same as that in Tsai & Wu (1992) when only the autocorrelation coefficient is of special interest.     

Influence Diagnostics in log-Birnbaum-Saunders Regression Models

In this paper we present various diagnostic methods for a linear regression model under a logarithmic Birnbaum-Saunders distribution for the errors, which may be applied for accelerated life testing or to compare the median lives of several populations. Some influence methods, such as the local influence, total local influence of an individual and generalized leverage are derived, analysed and discussed. We also present a connection between the local influence and generalized leverage methods. A discussion of the computation of the likelihood displacement as well as the normal curvature in the local influence method are presented. Finally, an example with real data is given for illustration.     

Local Influence Under Parameter Constraints

Calculations of local influence curvatures and leverage have been well developed when the parameters are unrestricted. In this article, we discuss the assessment of local influence and leverage under linear equality parameter constraints with extensions to inequality constraints. Using a penalized quadratic function we express the normal curvature of local influence for arbitrary perturbation schemes and the generalized leverage matrix in interpretable forms, which depend on restricted and unrestricted components. The results are quite general and can be applied in various statistical models. In particular, we derive the normal curvature under three useful perturbation schemes for generalized linear models. Four illustrative examples are analyzed by the methodology developed in the article.     

Influence diagnostics for polyhazard models in the presence of covariates

In this paper, we present various diagnostic methods for polyhazard models. Polyhazard models are a flexible family for fitting lifetime data. Their main advantage over the single hazard models, such as the Weibull and the log-logistic models, is to include a large amount of nonmonotone hazard shapes, as bathtub and multimodal curves. Some influence methods, such as the local influence and total local influence of an individual are derived, analyzed and discussed. A discussion of the computation of the likelihood displacement as well as the normal curvature in the local influence method are presented. Finally, an example with real data is given for illustration.     

Second-order inference for generalized least squares

Confidence regions for generalized least squares are commonly derived from a measure of departure calculated on the tangent plane at the MLE or on the tangent plane at the true value; the first gives approximate confidence regions, the second exact. For surfaces with curvature, indeed with varying curvature, the exact regions typically are not likelihood regions and can include parameter values of highest and of lowest likelihood. This paper develops an alternative approach to deriving exact confidence regions and uses both surface curvature and distance from the surface as supporting ingredients. For this, conditionality is invoked in two ways beyond that supported by the usual conditionality principle. For the case of normal error the ordinary chi-squared departure is replaced by a Von Mises-type angular (or cosine) departure which is assessed using curvature properties in the data direction and radial distance of the data from the regression surface. For the usual linear model (constant curvature equal to zero) the method coincides with the ordinary tests and confidence regions; for the case of constant nonzero curvature, the method generalizes to spheres and sphere-cylinders the Fisher (Statistical Methods and Scientific Inference, 1956) analysis of a rotationally symmetric normal on ?² with mean constrained to a circle. The effects of conditioning are indicated by a computer plot for obtaining 95% confidence.     

Local influence of nonlinear mixed effects model based on M-estimation

Abstract

In this work we mainly study the local influence in nonlinear mixed effects model with M-estimation. A robust method to obtain maximum likelihood estimates for parameters is presented, and the local influence of nonlinear mixed models based on robust estimation (M-estimation) by use of the curvature method is systematically discussed. The counting formulas of curvature for case weights perturbation, response variable perturbation and random error covariance perturbation are derived. Simulation studies are carried to access performance of the methods we proposed. We illustrate the diagnostics by an example presented in Davidian and Giltinan, which was analyzed under the non-robust situation.     

A note on influence diagnostics in AR(1) time series models

The purpose of this paper is to develop influence diagnostics for AR(1) models under the innovative and the data perturbation schemes. There are four main contributions. First, we derive analytical expressions for the slope and curvature statistics. Second, we establish a relationship between the slope and curvature showing that the standardised slope and standardised curvature are equal for the innovative perturbation scheme, and these vectors are nearly identical for several values of the autoregressive parameter, for the data perturbation scheme. Third, we present a connection between the influence statistics and the tests for outlier detection. Fourth, for the innovative perturbation scheme, we derive the asymptotic distribution of a new influence statistic, whereas for the data perturbation scheme, the distribution of the influence statistics is obtained via Monte Carlo simulation. We additionally discuss practical guidelines for the use of local influence statistics, which are illustrated on a chemical process data set.     

On Tests for Equicorrelation Coefficient of a Standard Symmetric Multivariate Normal Distribution

Sampson (1976, 1978) has considered applications of the standard symmetric multivariate normal (SSMN) distribution and the estimation of its equi-correlation coefficient, ρ. Tests for ρ are considered here. The likelihood ratio test suffers from several theoretical and practical shortcomings. We propose the locally most powerful (LMP) test which is globally (one-sided) unbiased, very simple to compute and is based on the best natural unbiased estimator of ρ. Exact null and non-null distributions of the test statistic are presented and percentage points are given. Statistical curvature (Efron, 1975) indicates that its performance improves with mk (sample size × dimension) while exact power computations show that even for reasonably small values of mk the performance is quite encouraging. Recalling Brown's (1971) cautions we establish by local comparison with the LMP similar test for ρ in the SMN (Rao, 1973) distribution, that here the additional information on the mean and variance is quite worthwhile.     

Sensitivity analysis for the identifiability with application to latent random effect model for the mixed data

In this paper, we study the indentifiability of a latent random effect model for the mixed correlated continuous and ordinal longitudinal responses. We derive conditions for the identifiability of the covariance parameters of the responses. Also, we proposed sensitivity analysis to investigate the perturbation from the non-identifiability of the covariance parameters, it is shown how one can use some elements of covariance structure. These elements associate conditions for identifiability of the covariance parameters of the responses. Influence of small perturbation of these elements on maximal normal curvature is also studied. The model is illustrated using medical data.     

A study of geometric and statistical curvatures for the skew-normal family of distributions

With special reference to the family of skew-normal distributions, we consider geometric curvature of a probability density function as a means to define and identify rare or catastrophic events—a phenomenon common in studying the financial instruments. Further, we study the statistical curvature properties of this family of distributions and discuss the sample size issue, to assess, to what extent the linear and likelihood-based inference of exponential family of distribution can be applicable for the skew-normal family.     

Diagnostics of Variance of the Error in Mixed Effects Linear Models Based on M-estimation

Fisher scoring method is applied to get M-estimator (robust estimator) of parameters in mixed effects linear models. Then influence curvature is used to study perturbation diagnostics of variance of the error based on M-estimation. The grape sugar data is used to illustrate the results.     

A new measure in local influence

The concept of local influence, based on the curvature of the log-likelihood surface and introduced by Cook in 1986, has suffered by not having well-established criteria for deciding whether individual cases are important in this respect or not. Here a new method for measuring the closeness of an individual case to the direction of maximum curvature is presented and discussed.     

Influence diagnostics in the tobit censored response model

In this article, we develop influence diagnostic tools for the tobit model. Specifically, we discuss global influence methods based on the Cook distance and residuals with envelopes, and total and conformal local influence techniques. In order to analyze the sensitivity of the maximum likelihood estimators of the parameters of the model to small perturbations on the assumptions of the model and/or data, we consider several perturbation schemes, such as case-weight and response perturbations. Finally, we illustrate the developed methodology by means of a real data set.     

Locally optimal test for simultaneously testing the mean and variance of a normal distribution

A generalization of the locally most powerful unbiased (LMPU) test for the single parameter case to the k-parameter case was proposed by SenGupta and Vermeire (1986). In particular we defined a locally most mean power unbiased (LMMPU) test based on the mean curvature of the power hypersurface. Compared to the type C tests of Neyman and Pearson and the type D tests (Isaacson, 1951), LMMPU tests possess better theoretical properties and enjoy ease of construction of critical regions. In this paper we present an interesting example of a two-parameter univariate normal population for which Isaacson (1951, p. 233) was unsuccessful in finding a type D test. For the case of one observation, we prove that no Type D region exists but the LMMPU test is obtained - it is an example of a test with singular Hessian matrix for its power but is nevertheless a strictly locally unbiased (LU) test.