Estimating probability of detection curves from regression data

Reliability assessments of nondestructive testing (NDT) techniques are rooted in statistical concepts. However, these concepts are sometimes overlooked, misapplied and misunderstood. Adding to the confusion is the multitude of approaches that often exist in the statistical literature for the same basic problem. This paper reviews fundamental concepts of using regression type data to establish probability of detection (POD) curves. The use of and some common misperceptions about confidence bounds are also discussed. Specific areas of discussion will include components of variation and confidence statements and how these concepts are related to the number of discontinuities in specimen sets and the number and nature of repeat tests.