A latent class method for the selection of prototypes using expert ratings.
Stat Med 2012 Jan; 30(1):80-92
Latent class analysis can be applied to the outcomes of expert ratings to select objects or subjects that are regarded as prototypical of a category in an ordinal classification system. During a pilot study, Monte Carlo simulations demonstrated that the probability of correct selection is larger when using latent class analysis than when usingmethods that rely on agreement statistics. Further improvements in the latent class results can also be achieved by applying affine transformations to latent class estimates of sensitivity and specificity. An application is presented that involves the selection of prototypical radiographs.
Statistical-analysis; Qualitative-analysis; Simulation-methods; Surveillance;
Author Keywords: Monte Carlo simulation; methodology; affine transformation; ordinal classification; random block design
William E. Miller, Division of Respiratory Disease Studies, National Institute for Occupational Safety and Health (NIOSH), Centers for Disease Control and Prevention (CDC), 1095 Willowdale Rd., Morgantown,WV 26505-2888, USA
Statistics in Medicine