1,721,020 research outputs found
The costruction of sampling weights in a regional survey on the willingness to pay for public or private long-term coverage
No full textComparing alternative distributional assumptions in mixed models used for small area estimation of income parameter
Full text linkLinear Mixed Models used in small area estimation usually rely on normality for the estimation of the variance components and the Mean Square Error of predictions. Nevertheless, normality is often inadequate when the target variable is income. For this reason, in this paper we consider Linear Mixed Models for the log-transformed income (which require back-transformation for prediction of means and totals on the variable’s original scale) and a Generalized Linear Mixed Model based on the Gamma distribution. Various prediction methods are compared by means of a simulation study based on the ECHP data. Standard predictors obtained from Linear Mixed Model for the untrasformed income are shown to be preferable to the considered alternatives, confirming their robustness with respect to the failure of the normality assumption
Small Area Estimation of Average Household Income Based on Panel Data
No full textThe availability of reliable estimates of income distribution parameters at a sub-national level is essential for the study of regional disparities. For European Union, countries the estimation of income parameters can be obtained from the European Community Household Panel, that provides reliable estimates for large areas within countries. The aim of this work is to find a suitable small area estimator of the average income at a smaller geographical scale, based on data from this survey. Since it is a panel survey, we suggest some different specifications of unit level Linear Mixed Models leading to small area estimators that borrow strength across both areas and times. In the Small Area Estimation context, full time series and “time series and cross-sectional” aggregate models have been applied in the literature, while small area unit level models for panel data have not been studied extensively. We compare the estimators performance by a Monte Carlo simulation study. Results show a significant gain in efficiency of estimators connected to models that take into account units autocorrelation
Measuring Sub-National Income Poverty by using a Small Area Multivariate approach
No full textThe present paper proposes a statistical strategy for the analysis of regional disparities in income
poverty. For the EU countries, information on individual income has been collected until now by the
European Community Household Panel survey, which only yields reliable estimates for very large
regions within countries. In order to obtain reliable estimates for some of the poverty indicators
suggested by the Laeken Council at the sub-national level, we suggest the adoption of a multivariate
small area estimation approach which enables us to reduce estimate variability. We concentrate on
Italy, the country with the lowest degree of regional cohesion within the EU. Results show that
disparity cannot be reduced to the so-called “North–South divide,” with the “poor” South separated
from the “affluent” North, as both these macro-regions display large internal differences in terms of
both poverty level and income inequality. The strategy we propose could also be adopted in order to
measure poverty in other European regions, using information produced by the new EU Survey on
Income and Living Conditions, which is replacing the European Community Household Panel
R package 'BayesLN'
No full textBayesian inference under log-normality assumption must be performed very carefully. In fact, under the common priors for the variance, useful quantities in the original data scale (like mean and quantiles) do not have posterior moments that are finite (Fabrizi et al. 2012 ). This package allows to easily carry out a proper Bayesian inferential procedure by fixing a suitable distribution (the generalized inverse Gaussian) as prior for the variance. Functions to estimate several kind of means (unconditional, conditional and conditional under a mixed model) and quantiles (unconditional and conditional) are provided
A comparison of Adjusted Bayes Estimators of an Ensemble of Small Area Parameters
No full textBayesian estimators of small area parameters may be very effective in improving the precision of “direct”, design-based estimates. By the way, the may be poor estimators of the actual Empirical Distribution Function of the ensemble of small area parameters. In this paper we review different adjusted estimators that correct or mitigate this problem in the context of Hierarchical Bayesian modelling, considering also extensions to the case of multivariate models that are widely used in Small Area estimation. In particular we consider constrained and linear constrained Hierarchical Bayes estimators, along with an estimation method recently by Zhang.
The posterior Mean Square Errors are proposed as measures of uncertainty. Different methods are compared by means of a simulation exercise in which many situations occurring in small area estimation applied problems are artificially re-created. The
main goals of the simulation is to assess the ability of various estimators to properly estimate the actual Empirical Distribution Function of area parameters and secondly to assess the frequentist properties of posterior MSEs. The method of Zhang emerges
as best in both the univariate and the multivariate setting, although the picture in the latter case is less clear
THE GENERALISED INVERSE GAUSSIAN DISTRIBUTION AS A PRIOR FOR DISEASE MAPPING MODELS
No full textIn this work we study sensitivity to hyperprior specification of the dispersion parameter in a convolution model widely used in spatial disease mapping, that accounts for both structured (spatial) and unstructured heterogeneity. In the fully Bayesian approach to disease mapping, hyperprior choice sensibly affects inferences. In this work we critically review the most common hyperprior specifications.
Moreover we propose a new hyperprior distribution, the Generalised Inverse Gaussian, starting from an idea explored in the estimation of the mean for iid log-Normal observations. In this context it is well known that hyperprior parameters have to be set accurately in order to avoid infinite moments of the posterior distribution. The performances of various hyperprior choices are compared via simulation studies. In particular we investigate the sensitivity of the relative risk estimates
Design Consistent Domain Level Estimation in Surveys with Massive Nonresponse
No full textRefusal of households in survey participation is becoming more frequent, and total nonresponse varies in subpopulations. Reweighting is performed in particular when information about nonrespondents is scarce. In survey practice, final sampling weights are defined in several stages: as the inverse of inclusion probabilities; then they are adjusted to reduce the bias in estimation due to unit nonresponse and undercoverage; eventually they are smoothed or trimmed, because the inflation of estimators' MSE is particularly severe at the domain level. In this paper, we deal with the design consistent estimation of domain parameters in surveys with massive nonresponse when the final sample is composed by two distinct parts drawn, with different designs, from two frames. We solve this problem of adapting a dual frame methodology by means of a simple model based on the Inverse Hypergeometric distribution. We also consider the smoothing of sampling weights. We propose a new trimming method, based on the Generalized Pareto distribution, which helps to improve the efficiency of estimators at the domain level
Multivariate Models for the Estimation of Poverty Indicators By Administrative Region and Household Type
No full textIn recent years, the measure of regional disparities on household poverty and social exclusion has received increasing attention by policy makers and this has produced a growing demand of sub-national statistical information on income parameters. Many studies have shown a strong connection between poverty and some characteristics of the household such as its composition, highlighting the concentration of the greatest economic difficulties in particular household typologies. For this reason, the aim of the work is to provide estimates of some poverty indicators for domains defined by cross-classifying the population by household typology (9 categories) and administrative region (20 categories), on the basis of data collected for Italy by the “European Union – Statistics on Income and Living Conditions” survey (EU-SILC). The need of estimating simultaneously more than one parameter is due to the multidimensionality of the studied phenomenon. In particular, we focus on three poverty rates based on three different thresholds, so to distinguish between poor people (PR, standard poverty threshold defined as the 60% of median of personal equivalent income), very poor people (PR80, threshold defined as the 80% of the standard poverty threshold) and people who are at risk of becoming poor (PR120, threshold defined as the 120% of the standard poverty threshold). The considered source, EU-SILC, is a sample survey on households’ income and social conditions, coordinated by Eurostat (Eurostat, 2005), designed to provide reliable estimates of main parameters of interest for areas within countries that do not correspond to our target domains. The number of units sampled from those domains are too scarce, in many cases, to obtain reliable estimates of the parameters of interest, therefore some small area estimation strategies have to be considered
- …
