From KIDSCREEN-10 to CHU9D: creating a unique mapping algorithm for application in economic evaluation

Table 2 Goodness-of-fit results from full sample

Model specification	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Mean CHU9D	Min CHU9D	Max CHU9D	MAE	RMSE	MAE^†	RMSE^†
Observed	0.8082	0.3479	1.0000	―	―	―	―
Method 1: Ordinary least squares estimator^‡
Model 1	0.8082	0.4535	0.9817	0.0978	0.1238	―	―
Model 2	0.8082	0.4909	1.0342	0.0950^**	0.1193^*	0.0946	0.1190
Method 2: Censored least absolute deviations estimator
Model 1	0.8185	0.4473	0.9944	0.0971	0.1243	―	―
Model 2	0.8179	0.4281	1.0802	0.0971	0.1247	0.0944	0.1219
Method 3: MM-estimator
Model 1	0.8136	0.4156	1.0019	0.0972	0.1243	0.0971	0.1243
Model 2	0.8146	0.4807	1.0555	0.0946^*	0.1199^**	0.0937	0.1193
Method 4: Generalised linear model
Model 1	0.8082	0.4693	0.9950	0.0975	0.1240	―	―
Model 2	0.8082	0.3760	0.9483	0.0971	0.1217	―	―

CHU9D – Child Health Utility 9D; MAE – mean absolute error; RMSE – root mean squared error.
*denotes the smallest value in the column; **denotes the second smallest value in the column.
^†The adjusted goodness-of-fit results by specifying the maximum predicted utility score to be 1.
^‡The R-square statistics for Model 1 and 2 are 0.36 and 0.41, respectively.

ISSN: 1477-7525