Maximum likelihood estimation of the prevalence of nonlethal neoplasms--an application to radon-daughter inhalation studies.