A frequent sampling design problem is length-biased sampling. This meansthat the probabilities of sample inclusion of population units are relatedto the values of the variable being measured. Parameter estimates maybecome biased and inconsistent if this sampling problem is ignored. Thispaper contains a review of applications characterized by length-biased samplingand the suggested solutions. The paper also includes a small simulationstudy on the properties of corrected mean estimators under misspecifiedsampling inclusion probabilities. Results indicate the importance ofcorrectly specified sampling inclusion mechanisms.
A commonly used sampling design in economic valuation studies is on-sitesampling. If this sampling design is used, the sampling inclusion probabil-ities may be correlated with respondents’ valuations, invalidating welfaremeasures derived from estimates of the probit model. This problem is re-ferred to a length-bias, a problem discovered in other fields of applicationof statistics.The first paper in this thesis outlines different application fields thathave length-bias problems and the suggested model solutions in the litera-ture are presented.The second paper of this thesis proposes a model based on the bivariateordinal probit, a model that can be used to analyze binary choice CV datagathered by on-site sampling. The models is presented, the log-likelihoodis derived, and the properties of the MLE’s are illustrated using a smallsimulation study. The simulation results show the proposed estimator tobe an interesting alternative.