Table 2.

Pros and cons of different ways of allocating resources to perform a study using 100 tests.

Subjects (n)	Tests per subject at different times	Tests per sample	Information on analytical imprecision (SDA and CVA)†	Information on within-subject biological variation (SDI and CVI)‡	Equation for SE of sampled population	Reduction of SE by this study design	Reduced dispersion of test result to improve classification§
100	1	1	No	No	${(\frac{{\mathit{SD}}_A^2+{\mathit{SD}}_I^2+{\mathit{SD}}_G^2}{100})}^{\frac{1}{2}}$	Most	None
50	2	1	No	Yes/limited‡	${([\frac{{\mathit{SD}}_A^2}{2}+\frac{{\mathit{SD}}_I^2}{2}+{\mathit{SD}}_G^2]/50)}^{\frac{1}{2}}$	Less	Yes
50	1	2	Yes†	No	${([\frac{{\mathit{SD}}_A^2}{2}+{\mathit{SD}}_I^2+{\mathit{SD}}_G^2]/50)}^{\frac{1}{2}}$	Less	Maybe
25	4	1	No	Yes/limited‡	${([\frac{{\mathit{SD}}_A^2}{4}+\frac{{\mathit{SD}}_I^2}{4}+{\mathit{SD}}_G^2]/25)}^{\frac{1}{2}}$	Less	Yes/best
25	2	2	Yes†	Yes†	${([\frac{{\mathit{SD}}_A^2}{4}+\frac{{\mathit{SD}}_I^2}{2}+{\mathit{SD}}_G^2]/25)}^{\frac{1}{2}}$	Least	Yes

It is assumed that 100 subjects are available and that performing duplicate measurements on individual samples and/or repeated measurements on individual subjects are options.

^†Analytical imprecision (SD_A and CV_A) can be estimated by taking the square root of the average variance of pairs of measurements, or other methods.

^‡The sum of within-subject variation (SD_I or CV_I) and analytical imprecision (SD_A or CV_A) can be estimated by taking the square root of the average variance of pairs of measurements separated in time; CV_I can only be accurately estimated if CV_A is determined separately and subtracted from (CV_I + CV_A).

^§Dispersion (confidence interval) for an individual test result is a function of CV_A and CV_I. Averaging repeated results separated in time will always reduce dispersion; averaging replicate results from each sample will only reduce dispersion if the magnitude of CV_A is similar to or greater than that of CV_I. How much misclassification is reduced depends on other factors (see text).

CV: Coefficient of variation; SD: Standard deviation; SD_G: Between-subject SD; SE: Standard error of the mean.