Estimation of victimization prevalence using data from the national crime survey (Q796953)

The author presents in this booklet a thorough overview of research work oriented to specification of victimization prevalence estimators for a large panel survey in the USA. In the beginning there is a description of sampling and rotation schemes and of the interview methods. In chapters two to four intuitive and model-based estimators are studied as to their sense, consistency and fit supposing no influence of missing data. For model based estimators the author offers both Bernoulli and Markov models of data. The estimators are that of maximum log-likelihood method. Chapter five deals with nonresponse, the sources being mentioned already in the introduction. The connection between nonresponse and victimization is discussed from several points of view and different ways of handling nonresponse are shown to lead to fairly different results. The last chapter of the paper offers some suggestions for further research and discussion. It should be stressed that all the estimation is done on ''housing unit level'', housing units being proper sampling units. Problems connected with transitions to ''household level'' (one housing unit may contain several households) are only roughly sketched. In the text the readers find a lot of concrete results and at the end of the book the questionnaire and some programs used for model fitting and estimation (in Fortran).