Abstract: Classifyingan individual unit into normal (N)
or abnormalcategories
will often be expensive and time consuming. Instead, if a screening variable,
which can be easily measured is available, one may use the screening variable
measurement to classify that individual. Assuming that the individual units in
the abnormal category takes smaller values of the screening variable, we need to
find a threshold u such that units
with screening variable measurement at most u
are classified and others are
classified N. In this paper, we
determine u so that the asymptotic
relative efficiency in testing the prevalence rate of with and without
screening variable measurement is maximum.
Keywords and phrases: false positive, false negative, normal distribution.