Premium
False positive rates in the determination of changes in probing depth‐related periodontal measurements
Author(s) -
Cohen Mark E.,
Ralls Stephen A.
Publication year - 1988
Publication title -
journal of periodontal research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.31
H-Index - 83
eISSN - 1600-0765
pISSN - 0022-3484
DOI - 10.1111/j.1600-0765.1988.tb01351.x
Subject(s) - false positive paradox , statistics , replicate , mathematics , magnitude (astronomy) , false positive rate , type i and type ii errors , periodontal disease , observational error , normal distribution , distribution (mathematics) , dentistry , medicine , physics , mathematical analysis , astronomy
False positive rates associated with changes in periodontal probing measurements (changes which are of such magnitude as to be construed as due to disease or healing when the observed changes are actually due to measurement error) were estimated by computerized simulation. In the first phase of the simulation study, various distributions of error variances among sites were evaluated for their ability to produce matches to an empirical distribution of differences between replicate measurements. In the second phase of the study, distributions of variances identified in Phase I were used to estimate the false positive rate, under conditions of no actual change, for detection methods based on critical differences between averaged pairs of measurements. This rate was found to be substantially greater than that predicted using normal distribution probabilities and, for a difference of ≥2.5 mm, approached one false detection per examination of 168 sites. In the third phase of the study, simulation procedures were extended to the tolerance detection methodology and the false positive rate, in the absence of real change, was almost one detection per two examinations. This simulation suggested that perhaps one third of tolerance detected “bursts” of periodontal attachment change may be false positives attributable to measurement error.