Abstract
The R package emdi enables the estimation of regionally disaggregated indicators using small area estimation methods and includes tools for processing, assessing, and presenting the results. The mean of the target variable, the quantiles of its distribution, the headcount ratio, the poverty gap, the Gini coefficient, the quintile share ratio, and customized indicators are estimated using direct and model-based estimation with the empirical best predictor (Molina and Rao 2010). The user is assisted by automatic estimation of datadriven transformation parameters. Parametric and semi-parametric, wild bootstrap for mean squared error estimation are implemented with the latter offering protection against possible misspecification of the error distribution. Tools for (a) customized parallel computing, (b) model diagnostic analyses, (c) creating high quality maps and (d) exporting the results to Excel and OpenDocument Spreadsheets are included. The functionality of the package is illustrated with example data sets for estimating the Gini coefficient and median income for districts in Austria.
Highlights
In recent years an increased number of policy decisions has been based on statistical information derived from indicators estimated at disaggregated geographical levels using small area estimation methods
The headcount ratio (HCR) describes the proportion of the population below the poverty line and the poverty gap (PG) further takes into account how far, on average, this proportion falls below the threshold
In this paper we show how the emdi package can simplify the application of small area estimation (SAE) methods
Summary
In recent years an increased number of policy decisions has been based on statistical information derived from indicators estimated at disaggregated geographical levels using small area estimation methods. More recently widespread application of SAE methods involves the estimation of poverty and inequality indicators and distribution functions (The World Bank 2007). In this case the use of methodologies for estimating means and totals is no longer appropriate since such indicators are complex, non-linear functions of the data. The sae package implements a range of small area methods, it lacks the necessary functionality for supporting the user beyond estimation for example, for performing model diagnostic analyses, visualizing, and exporting the results for further processing.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have