Seminar Notice

R Seminar Notice


            Some Longitudinal Data Issues in Using the 
              National Crime Victimization Survey
  
                     Sharon Lohr
                Department of Mathematics
                 Arizona State University
  
We briefly describe the design and use of the U.S. National Crime 
Victimization Survey (NCVS) for estimating national crime victimization
rates.  Victimizations are not randomly scattered through the population,
however, but tend to be concentrated in relatively few victims.
We estimate the conditional probabilities of being a crime victim at time to 
given the victimization status in earlier interviews.  One challenge to
estimating the conditional probabilities is that the NCVS follows housing
units, not households, so the subsequent victimization experience of
households that move partway through the seven interviews is not available.
A model using available data for a scheduled housing unit is
presented to allow use of partial information from households that
move in or out of the housing unit during the study period.
Possible implications for estimating cross-sectional victimization
rates are discussed.