Data used to calibrate rating curve models commonly contain values reported
as ``less-than'' an analytical detection limit.
Cohn et al. [1992b] and
Cohn [1988]
report an adjusted maximum likelihood estimator, 
, for
use with data
sets containing such ``less-than'' values.
is a generalization of 
, and was found to
share many of its desirable properties:
is nearly-unbiased, and the variance
of sums of loads is easy to compute.