Fig. 3From: Using linked administrative data to aid the handling of non-response and restore sample representativeness in cohort studies: the 1958 national child development study and hospital episode statistics dataMean (95% confidence interval) cognitive ability at age 7 in the National Child Development Study before and after handling missing data (Analysis B)Analysis 1: Distribution using all available data (n = 14,407)Analysis 2: Distribution restricted to wave 9 target population (alive and still living in Great Britain) (n = 12,938)Analysis 3: Distribution restricted to wave 9 respondents (n = 7,839)Analysis 4: MI analysis using only selected survey predictors of non-response as auxiliary variables, restricted to wave 9 target population and those non-missing for the variable of interest, using information on the variable of interest from wave 9 respondents only (n = 12,938)Analysis 5: MI analysis using selected HES predictors of non-response in addition to selected survey predictors of non-response as auxiliary variables, restricted to wave 9 target population and those non-missing for the variable of interest, using information on the variable of interest from wave 9 respondents only (n = 12,938)Back to article page