Mariangela Sciandra

Estimating growth charts via nonparametric quantile regression: a practical framework with application in ecology

We discuss a practical and effective framework to estimate reference growth charts via regression quantiles. Inequality constraints are used to ensure both monotonicity and non-crossing of the estimated quantile curves and penalized splines are employed to model the nonlinear growth patterns with respect to age. A companion R package is presented and relevant code discussed to favour spreading and application of the proposed methods.

Quantile regression via iterative least squares computations

We present an estimating framework for quantile regression where the usual L 1-norm objective function is replaced by its smooth parametric approximation. An exact path-following algorithm is derived, leading to the well-known ‘basic’ solutions interpolating exactly a number of observations equal to the number of parameters being estimated. We discuss briefly possible practical implications of the proposed approach, such as early stopping for large data sets, confidence intervals, and additional topics for future research.

Subject-specific odds ratios in binomial GLMMs with continuous response

In a regression context, the dichotomization of a continuous outcome variable is often motivated by the need to express results in terms of the odds ratio, as a measure of association between the response and one or more risk factors. Starting from the recent work of Moser and Coombs (Stat Med 23:1843–1860, 2004) in this article we explore in a mixed model framework the possibility of obtaining odds ratio estimates from a regression linear model without the need of dichotomizing the response variable. It is shown that the odds ratio estimators derived from a linear mixed model outperform those from a binomial generalized linear mixed model, especially when the data exhibit high levels of heterogeneity.

A PCA Interpretation of the Glasgow Coma Scale in the Trauma Brain Injury PECARN Dataset

CT scan is strongly recommended for a patient affected by head trauma, but he/she must absorb a certain amount of radiations. For this reason, the physician tries to avoid such a practice for pediatric patients. The symptoms analysis, visual/tactile inspection, and reactions to appropriate stimuli from the physician could induce him/her to put the patient in a period of observation instead of performing an immediate CT scan. As a consequence, the correct evaluation of those symptoms is a crucial task. For this reason, the Pediatric Glasgow Coma Scale (PGCS) plays a fundamental role, because it is a numeric scale regarding the patient’s mental status. It is computed as the sum of the score for the eye, motor and verbal response evaluated by the physician. In this paper, the Principal Component Analysis (PCA) is performed on the PGCS of the Trauma Brain Injury (TBI) dataset collected by the PECARN (Pediatric Emergency Care Applied Research Network). The PCA is performed in all cases when the sum of the three partial scores results in a value less than 14, because a patient with PGCS = 15 is not considered at risk. Under this constraint, the PCA reveals that each partial GCS give the same contribution to the first principal component, but a scale variation is introduced.

New Flexible Probability Distributions for Ranking Data

Recently, several models have been proposed for analysing the ranks assigned by people to some object. These models summarize the liking feeling towards the object, possibly with respect to a set of explanatory variables. Some recent works have suggested the use of the Shifted Binomial and of the Inverse Hypergeometric distribution for modelling the approval rate, while mixture models have been considered for taking into account the uncertainty in the ranking process. We propose two new probability distributions, the Discrete Beta and the Shifted-Beta Binomial, which ensure much flexibility and allow the joint modelling of the scale (approval rate) and the shape (uncertainty) parameters of the rank distribution.