Diagnosing problems with imputation models using the Kolmogorov-Smirnov test: a simulation study

Table 1 Linear regression analysis results for the Longitudinal Study of Australian Children example

Variable	Missing	Complete case analysis		Model 1		Model 2		Model 3		Model 4
Variable	n (%)	Complete case analysis		Model 1		Model 2		Model 3		Model 4
		Coefficient	SE	Coefficient	SE	Coefficient	SE	Coefficient	SE	Coefficient	SE
Mother completed high school	3 (0.1)	−0.147	0.061	−0.216	0.054	−0.212	0.054	−0.233	0.054	−0.212	0.053
Socioeconomic position	151 (3.6)	−0.192	0.028	−0.183	0.026	−0.184	0.026	−0.184	0.025	−0.180	0.025
Male child	0 (0)	0.236	0.048	0.281	0.043	0.283	0.043	0.284	0.043	0.281	0.043
Warm parenting	251 (6.0)	−0.309	0.059	−0.297	0.053	−0.290	0.055	−0.295	0.054	−0.303	0.054
Harsh discipline score	970 (23.0)	0.252	0.017	0.232	0.018	0.212	0.017	0.247	0.017	0.241	0.018
K6 total score	287 (6.8)	0.033	0.008	0.035	0.007	0.036	0.008	0.032	0.007	0.033	0.008
Smoker	971 (23.1)	0.155	0.069	0.193	0.063	0.199	0.062	0.153	0.077	0.208	0.063

The multiple imputation results were estimated using 20 imputations. The four imputation models were: 1 = “optimal” model, 2 = no outcome variable, 3 = no auxiliary variables, 4 = no de-skewing.

ISSN: 1471-2288