Alfio Marazzi - Academia.edu (original) (raw)

Papers by Alfio Marazzi

Robust estimators for generalized linear models

Journal of Statistical Planning and Inference, 2014

Abstract In this paper we propose a family of robust estimators for generalized linear models. Th... more Abstract In this paper we propose a family of robust estimators for generalized linear models. The basic idea is to use an M-estimator after applying a variance stabilizing transformation to the response. We show the consistency and asymptotic normality of these estimators. We also obtain a lower bound for their breakdown point. A Monte Carlo study shows that the proposed estimators compare favorably with respect to other robust estimators for generalized linear models with Poisson response and log link.

Variation des durées et ou des coûts réels d'hospitalisation au sein des DRG individuels : revue de la littérature sur l'existence du phénomène et ses causes possibles

Running head Truncating Length of Stay

Most distributions of hospital length of stay are asymmetric, with a long right tail and some ver... more Most distributions of hospital length of stay are asymmetric, with a long right tail and some very large observations (outliers). These features vitiate the reliability of many statistical summaries, such as the arithmetic mean, and comparisons based on them. A common remedy is to truncate (i.e., remove) values outside some limits and take the arithmetic mean of the remaining values. In general, the limits are based on a position measure (e.g., mean, median, quartiles) and a scale measure (e.g., standard deviation, median absolute deviation, interquartile range). In addition, a scale transformation (usually the logarithm) is frequently used. Using a data base with almost five millions hospital stays from five European countries, this paper explores the performance of five common truncation rules combining various options on transformation, position and scale. These rules are compared with a new one called « approximated quartile based truncated mean » or AQTM. The AQTM is based on a...

Statistics & Risk Modeling, 1985

The restricted Bayes and minimax principles of Hodges and Lehmann [8] are applied to the problem ... more The restricted Bayes and minimax principles of Hodges and Lehmann [8] are applied to the problem of estimating the parameters of a linear model when : a) the error distribution is Gaussian and the prior distribution is not exactly known; b) the prior distribution is Gaussian and the given error distribution is not precise. Approximate analytical and numerical solutions are studied. robustness problems were proposed by Hodges and Lehmann [8].

A note on estimating a mean cost of hospital stay with incomplete information

Consistency of the robust residual autocorrelation estimate of a transformation parameter, 2005

ABSTRACT AMS classification. Primary 62J05; secondary 62F35. Abstract. The linear regression mode... more ABSTRACT AMS classification. Primary 62J05; secondary 62F35. Abstract. The linear regression models for a transformed response is consid- ered. S- and MM-estimates depending on the transformation parameter λ are defined and asymptotic results for these estimates are obtained. Using these results, consistency of the robust residual autocorrelation estimate of λ based on S- and MM-estimates is proved in the simple regression case.

Demographic Research a free, expedited, online journal of peer-reviewed research and commentary i... more Demographic Research a free, expedited, online journal of peer-reviewed research and commentary in the population sciences published by the

Journal of Statistical Software, 2016

robustloggamma is an R package for robust estimation and inference in the generalized loggamma mo... more robustloggamma is an R package for robust estimation and inference in the generalized loggamma model. We briefly introduce the model, the estimation procedures and the computational algorithms. Then, we illustrate the use of the package with the help of a real data set.

Stats, 2021

The distance constrained maximum likelihood procedure (DCML) optimally combines a robust estimato... more The distance constrained maximum likelihood procedure (DCML) optimally combines a robust estimator with the maximum likelihood estimator with the purpose of improving its small sample efficiency while preserving a good robustness level. It has been published for the linear model and is now extended to the GLM. Monte Carlo experiments are used to explore the performance of this extension in the Poisson regression case. Several published robust candidates for the DCML are compared; the modified conditional maximum likelihood estimator starting with a very robust minimum density power divergence estimator is selected as the best candidate. It is shown empirically that the DCML remarkably improves its small sample efficiency without loss of robustness. An example using real hospital length of stay data fitted by the negative binomial regression model is discussed.

Robust Gamma regression models for the analysis of health care cost data

Model Assisted Statistics and Applications, 2012

The population-mean cost of patients with certain pathologies is the parameter of interest for al... more The population-mean cost of patients with certain pathologies is the parameter of interest for allocating health resources. It generally depends upon a number of covariates and the presence of outliers yields difficulties in the estimation procedure. Recent research in parametric robust techniques proposed the use of robust estimating equations via M-estimation for the Gamma model [2] and a class of high efficiency and high breakdown point estimators [14] extended to the case of generalized log-gamma

Handouts for the Instructional Meeting on "Robust Statistical Methods

Using past experience to optimize audit sampling design

Review of Quantitative Finance and Accounting, 2016

Computational Statistics & Data Analysis, 2017

The generalized log-gamma (GLG) model is a very flexible family of distributions to analyze datas... more The generalized log-gamma (GLG) model is a very flexible family of distributions to analyze datasets in many different areas of science and technology. In this paper, we propose estimators which are simultaneously highly robust and highly efficient for the parameters of a GLG distribution in the presence of censoring. We also introduced estimators with the same properties for accelerated failure time models with censored observations and error distribution belonging to the GLG family. We prove that the proposed estimators are asymptotically fully efficient and examine the maximum mean square error using Monte Carlo simulations. The simulations confirm that the proposed estimators are highly robust and highly efficient for finite sample size. Finally, we illustrate the good behavior of the proposed estimators with two real datasets.

A robust conditional maximum likelihood estimator for generalized linear models with a dispersion parameter

TEST

Robust Estimators of the Generalized Long-Gamma Distribution

Technometrics a Journal of Statistics For the Physical Chemical and Engineering Sciences, 2014

Restricted minimax credibility: Two special cases

Sozial Und Praaventivmedizin Spm, Feb 1, 1993

Le D partement de Statistique de I'Institut Universitaire de M6decine Sociale et Preventive, Lausanne

Soz Praventivmed, 1982

Robust Bayesian estimation for the linear model