In Reply.—
Acquisition of numerical data from any graphic display, including dot plots, implies some imprecision. Thus, the number of subjects within each result interval might vary slightly on different runs. The "correct" values should reflect original data, not data acquired by any particular method or individual. Occasionally, electronic "noise" in data tablets might inject "extra points," which are easily detected and eliminated by comparison with original documents.Data acquisition from dot diagrams typically has an imprecision of about 3%,1 comparable with imprecision from preanalytic variation, analytic variation, and interlaboratory variation.We agree that reports on test performance should provide tables of numerical data with multiple test results over multiple clinical states. However, when such tables are not available, dot plots can provide numerical data sufficiently accurate for clinical use. From this numerical data, using spread-sheet software for desktop computers, many evaluations can be performed, including cumulative frequency distributions,