Imputation, modelling or weighting shall be applied to the data where necessary.
Where non-response to income variables at component level results in missing data, appropriate methods of statistical imputation shall be applied.
Where any net income variable at component level is collected directly, appropriate methods of statistical imputation or modelling, or both, shall be applied to obtain the required target gross variables and vice versa.
Where non-response to an individual questionnaire occurs within a sample household, appropriate statistical methods for statistical weighting or imputation shall be used to estimate the total income of the household.
The procedure applied to the data shall preserve the variation in and the correlation between variables. Methods that incorporate ‘error components’ into the imputed values shall be preferable to those that simply impute a predicted value.
Methods which take into account the correlation structure (or other characteristics of the joint distribution of the variables) shall be preferable to the marginal or univariate approach.