As described in the previous columns (1,2), confusion can be present in multidisciplinary uses of statistical methods for
analytical comparisons due to the variation in terminology, assumptions, and specific uses within different technical disciplines.
Certain analytical methods of analysis are concerned only with overall error of analysis or correlation, others emphasize
repeatability and reproducibility, and yet others report only bias and precision. Often, engineers might look at total error
analysis as tolerance stacking so that the sums of all errors in a specific dimension are less than the total maximum allowable
tolerance.

Jerome Workman, Jr.

As noted in Part I, a unification of multiple disciplines into a reasonable set of statistical parameters remains useful for
a broad presentation of clinical analytical data (1). A single set of statistical parameters is useful to a multidisciplinary
team involved in looking at analytical method comparison.

Table I: Sample data (sample calculations use data from reference 5 for comparison)

In Part II, we described the use of an Excel spreadsheet that computes and displays the statistical parameters described in
this series (2). For each parameter, the basic Excel computation nomenclature is described. The equations and terminology
in this paper are consistent with Clinical and Laboratory Standards Institute (CLSI) guidelines (3); please refer specifically
to these guidelines referenced below for your own assessment. These statistical analyses evaluate the accuracy of a test method
compared to a reference method measuring the same analyte.

Figure 1

Refer to the previous Part II column for more detailed definitions of terms and to references 4–6 for additional descriptions
and worked problems associated with the individual statistics demonstrated in this article. Note the data used for Parts I
and II are shown in Table I and Figure 1. Note the entire Excel spreadsheet with computed results is shown in Figure 2, and
the specific locations of data and calculations by rows and columns using Excel computational formulas are given in Part II.

Figure 2

Table I shows a set of reference values and a set of test values for seven samples; note that the data in Table I are shown
in Excel format in Figure 1. The regression plot is generated in Figure 3a showing the reference (X) data and the test (Y) data with a regression line, the slope and intercept of the regression line, and the coefficient of determination (or R^{2}); the residual plot is shown in Figure 4. The specific Excel functions for making these plots are given for plotting the
X–Y data (Figures 3b–3d) and for the regression plot (Figures 3b–3f). In addition, various graphs are illustrated for correcting
one analytical method to report the values of a second method (Figures 5 and 6). Plots commonly used for comparison of two
analytical methods when no "gold" reference method is available are given in Table II and in Figures 7 and 8.