Thin plate regression splines

Summary. I discuss the production of low rank smoothers for d  ≥ 1 dimensional data, which can be fitted by regression or penalized regression methods. The smoothers are constructed by a simple transformation and truncation of the basis that arises from the solution of the thin plate spline smoothing problem and are optimal in the sense that the truncation is designed to result in the minimum possible perturbation of the thin plate spline smoothing problem given the dimension of the basis used to construct the smoother. By making use of Lanczos iteration the basis change and truncation are computationally efficient. The smoothers allow the use of approximate thin plate spline models with large data sets, avoid the problems that are associated with 'knot placement' that usually complicate modelling with regression splines or penalized regression splines, provide a sensible way of modelling interaction terms in generalized additive models, provide low rank approximations to generalized smoothing spline models, appropriate for use with large data sets, provide a means for incorporating smooth functions of more than one variable into non-linear models and improve the computational efficiency of penalized likelihood models incorporating thin plate splines. Given that the approach produces spline-like models with a sparse basis, it also provides a natural way of incorporating unpenalized spline-like terms in linear and generalized linear models, and these can be treated just like any other model terms from the point of view of model selection, inference and diagnostics.     

Double hierarchical generalized linear models (with discussion)

Summary.  We propose a class of double hierarchical generalized linear models in which random effects can be specified for both the mean and dispersion. Heteroscedasticity between clusters can be modelled by introducing random effects in the dispersion model, as is heterogeneity between clusters in the mean model. This class will, among other things, enable models with heavy-tailed distributions to be explored, providing robust estimation against outliers. The h -likelihood provides a unified framework for this new class of models and gives a single algorithm for fitting all members of the class. This algorithm does not require quadrature or prior probabilities.     

On the distribution of crossings in a generalized random walk

This paper deals with the derivation of the probability distribution of the rank order statistic N¹_μ,n(r), the number of crossings of heightr(≥0) in a generalized random walk with steps 1 and ?μ by using the modified Dwass technique.     

Lee–Carter mortality forecasting: a parallel generalized linear modelling approach for England and Wales mortality projections

Summary. The paper presents a reinterpretation of the model underpinning the Lee–Carter methodology for forecasting mortality (and other vital) rates. A parallel methodology based on generalized linear modelling is introduced. The use of residual plots is proposed for both methods to aid the assessment of the goodness of fit. The two methods are compared in terms of structure and assumptions. They are then compared through an analysis of the gender- and age-specific mortality rates for England and Wales over the period 1950–1998 and through a consideration of the forecasts generated by the two methods. The paper also compares different approaches to the forecasting of life expectancy and considers the effectiveness of the Coale–Guo method for extrapolating mortality rates to the oldest ages.     

The Equality of the Ordinary Least Squares Estimator and the Best Linear Unbiased Estimator

It is well known that the ordinary least squares estimator of Xβ in the general linear model E _y = Xβ, cov y = σ² V, can be the best linear unbiased estimator even if V is not a multiple of the identity matrix. This article presents, in a historical perspective, the development of the several conditions for the ordinary least squares estimator to be best linear unbiased. Various characterizations of these conditions, using generalized inverses and orthogonal projectors, along with several examples, are also given. In addition, a complete set of references is provided.     

Calculation of Scores for a Wilcoxon Generalization Applicable to Data Subject to Arbitrary Right Censorship

For data subject to right censoring it is suggested that the Wilcoxon ranking procedure can be generalized by scoring observations according to the expected values of order statistics from the uniform distribution subject to the same right censoring. This parallels the logrank scoring procedure in which scores correspond to the expected values of order statistics from the exponential distribution that have been subject to right censoring. A caveat is given that, in situations where the mechanism of censoring has been affected by treatment, the usual permutational analysis of ranking scores would be inappropriate. But a jackknife approach could be remedial.     

Upper Bounds on the True Coverage of Bootstrap Percentile Type Confidence Intervals

A simple derivation of expected mean squares is given for the randomized (complete) block design, showing that “experimental error,” the error term for testing treatments, is comprised of three sources of variability: block by treatment interaction, within block plot-to-plot variability, and within experimental plot sampling variation. The approach could readily be extended to incorporate measurement error as a fourth component of experimental error.     

On the Properties of the SAVE Directions

In quadratic discriminant analysis, the use of SAVE (Cook and Weisberg, 1991 Cook, R.D., Weisberg, S. (1991). Discussion of Li (1991). J. Amer. Statist. Assoc. 86:328–32.[Taylor &; Francis Online], [Web of Science ®] , [Google Scholar]; Pardoe et al., 2007 Pardoe, I., Yin, X., Cook, R. (2007). Graphical tools for quadratic discriminant analysis. Technometrics 49:172–183.[Taylor &; Francis Online], [Web of Science ®] , [Google Scholar]) is often recommended for dimension-reduction purposes. However, the associated directions tend to over-emphasize the differences of the groups in dispersion, ignoring at the same time those in location. This behavior makes often the plots of the corresponding canonical coordinates difficult to interpret. In this article, the properties of SAVE are investigated and related to those of the SIR and SIRII components. Applications with real data are presented. Comparisons with previous work in this area are also discussed.     

Random Effects Modeling and the Zero-Inflated Poisson Distribution

Overdispersion due to a large proportion of zero observations in data sets is a common occurrence in many applications of many fields of research; we consider such scenarios in count panel (longitudinal) data. A well-known and widely implemented technique for handling such data is that of random effects modeling, which addresses the serial correlation inherent in panel data, as well as overdispersion. To deal with the excess zeros, a zero-inflated Poisson distribution has come to be canonical, which relaxes the equal mean-variance specification of a traditional Poisson model and allows for the larger variance characteristic of overdispersed data. A natural proposal then to approach count panel data with overdispersion due to excess zeros is to combine these two methodologies, deriving a likelihood from the resulting conditional probability. In performing simulation studies, we find that this approach in fact poses problems of identifiability. In this article, we construct and explain in full detail why a model obtained from the marriage of two classical and well-established techniques is unidentifiable and provide results of simulation studies demonstrating this effect. A discussion on alternative methodologies to resolve the problem is provided in the conclusion.