|
Imputation in the survey on living conditions |
José María Méndez Martín
Instituto Nacional de Estadística
Esta dirección de correo electrónico está protegida contra los robots de spam, necesita tener Javascript activado para poder verla
- Abstract
The Spanish "European Statistics on Income and Living Conditions" (EU-SILC) is one of the statistical operations that has been harmonised to EU standards. In this household survey there are two kinds of non-response: unit non-response (one or several household or individual questionnaires are missing) and item non-response (no questionaire is missing but some variables are). The main target variable is the total household income, which is defined as the aggregate of different income components. Components with missing values are imputed when they cannot be estimated with the help of other variables or other information in the questionnaire of the current or previous surveys. The procedure applied to the data preserves the variability of the variables and the correlations between them. The statistical software used for imputation is the IVEware. The IVEware implements a multivariate model involving a multiple regression sequence where imputation is carried out variable by variable generating draws from the predictive distribution specified by the regression model. An iterative imputation scheme is used, updating previous imputed values in order to better preserve the correlation among variables.
Keywords: Risk EU-SILC, imputation, item non-response, IVEware.
 PDF (177.26 KB)
|
