Table 2.

Computations made with a series of 10 EBCR onset latency values (ms).

Method	Untransformed data		Box-Cox transformed data		Nonparametric
	Standard	Robust	Standard	Robust
N	10	10	10	10	10
Mean	18.72		1.47
Median	18.40	18.57	1.47	1.47
SD	2.69	2.87	0.06	0.06
Minimum	15.2	15.2	1.39	1.39
Maximum	23.5	23.5	1.56	1.56
λ 1 coefficient Box-Cox			−5.185	−5.185
λ 2 coefficient Box-Cox			−0.484	−0.484
P value Anderson–Darling	0.685		0.923
Symmetry test for robust		0.746		0.998
Outliers Dixon
Outliers Tukey	0	0	0	0
Suspect data Tukey	0	0	0	0
Lower limit of reference interval	12.34a	12.07a	13.92a	13.85a
Upper limit of the reference interval	25.10a	25.07a	27.30a	27.44a
90% CI for lower limit	9.98		12.65
	15.10		15.59
90% CI for upper limit	22.22		22.46
	27.81		34.50
Comments
The sample size is too small (n < 40) to compute a nonparametric reference interval. Data should be analyzed with the nonparametric method. As an alternative, the robust method with a Box-Cox transformation may be used after checking the symmetry of the distribution. The robust method with a Box-Cox transformation of the data gives the following reference interval: [13.85; 27.44]. The sample size is too small (n < 20) to compute the CIs for the limits of the reference interval obtained with the robust method with Box-Cox transformation.

^aIn agreement with the IFCC-CLSI recommendation.