A multistep approach for imputation of missing values in composite indicators
Although they have been selected on the basis of their reliability, consistency, continuity and completeness, most of indicators used in INFORM Global Risk Index do not have global coverage and neither are issued regularly every year. This results in a significant number of missing values, irregularly distributed among countries, time and indicators.
The main motivations for imputing missing values arise from the need to create consistent trends that would otherwise not be possible due to the lack of data in the indicator’s time series, and to increase the reliability of the single compound release.
In the presented study we focus on better understanding the patterns and mechanisms of missing values in the INFORM GRI model, and on evaluating their impact on the model’s outputs. The scope is to develop a missing data imputation strategy to be implemented in the INFORM GRI that will strongly depend on the reason why data is missing.
POLJANSEK Karmen;
VERNACCINI Luca;
NWEKE Eje;
MARIN FERRER Montserrat;
2020-01-09
Publications Office of the European Union
JRC119496
978-92-76-14657-5 (online),
978-92-76-14725-1 (print),
1831-9424 (online),
1018-5593 (print),
EUR 30037 EN,
OP KJ-NA-30037-EN-N (online),
OP KJ-NA-30037-EN-C (print),
https://publications.jrc.ec.europa.eu/repository/handle/JRC119496,
10.2760/97910 (online),
10.2760/717316 (print),
