arXiv subjects

Andrea Saltelli

Publications and source records attributed to Andrea Saltelli.

22 records · Page 2Linked to original sources

arXiv2015

Exploring multi-dimensional spaces: a Comparison of Latin Hypercube and Quasi Monte Carlo Sampling Techniques

Three sampling methods are compared for efficiency on a number of test problems of various complexity for which analytic quadratures are available. The methods compared are Monte Carlo with pseudo-random numbers, Latin Hypercube Sampling, and Quasi Monte Carlo with sampling based on Sobol sequences. Generally results show superior performance of the Quasi Monte Carlo approach based on Sobol sequences in line with theoretical predictions. Latin Hypercube Sampling can be more efficient than both Monte Carlo method and Quasi Monte Carlo method but the latter inequality holds for a reduced set of function typology and at small number of sampled points. In conclusion Quasi Monte Carlo method would appear the safest bet when integrating functions of unknown typology.

stat.AP↗

arXiv2014

Exploring Hoover and Perez's experimental designs using global sensitivity analysis

This paper investigates variable-selection procedures in regression that make use of global sensitivity analysis. The approach is combined with existing algorithms and it is applied to the time series regression designs proposed by Hoover and Perez. A comparison of an algorithm employing global sensitivity analysis and the (optimized) algorithm of Hoover and Perez shows that the former significantly improves the recovery rates of original specifications.

stat.CO↗

arXiv2012

What do I make of your Latinorum? Sensitivity auditing of mathematical modelling

In their book Uncertainty and Quality of Science for Policy (1990), Funtowicz and Ravetz argued the need to extend traditional methods and techniques of quality assurance in policy-related science. Since then, these ideas have been operationalized further and applied. Particularly relevant have been the recourse to extended peer review, to be intended as internal, across disciplines, as well as external, between practitioners and stakeholders, and the use of a new approach to qualify quantities: NUSAP (Numeral, Unit, Spread, Assessment, Pedigree). Here we describe how sensitivity analysis, mandated by existing guidelines as a good practice to use in conjunction to mathematical modelling, needs to be transformed and adapted to ensure quality in the treatment of uncertainty of science for policy. We thus provide seven rules to extend the use of sensitivity analysis (or how to apportion uncertainty in model based inference among input factors) in a process of sensitivity auditing of models used in a policy context. Each rule will be illustrated by examples.

physics.soc-ph↗

arXiv2012

Ratings and rankings: Voodoo or Science?

Composite indicators aggregate a set of variables using weights which are understood to reflect the variables' importance in the index. In this paper we propose to measure the importance of a given variable within existing composite indicators via Karl Pearson's `correlation ratio'; we call this measure `main effect'. Because socio-economic variables are heteroskedastic and correlated, (relative) nominal weights are hardly ever found to match (relative) main effects; we propose to summarize their discrepancy with a divergence measure. We further discuss to what extent the mapping from nominal weights to main effects can be inverted. This analysis is applied to five composite indicators, including the Human Development Index and two popular league tables of university performance. It is found that in many cases the declared importance of single indicators and their main effect are very different, and that the data correlation structure often prevents developers from obtaining the stated importance, even when modifying the nominal weights in the set of nonnegative numbers with unit sum.

stat.AP↗