Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


Thesis_Hui.pdf (1.64 MB)

ETD Abstract Container

Abstract Header

Adjusting for Bounding and Time-in-Sample E ects in the National Crime Victimization Survey (NCVS) Property Crime Rate Estimation

Yang, Hui
http://rave.ohiolink.edu/etdc/view?acc_num=osu1452167047

Abstract Details

2016, Doctor of Philosophy, Ohio State University, Statistics.
In this study, we deal with two problems: rotation group bias and lack of bounding information for producing crime rate estimates using the National Crime Victimization Survey (NCVS) data. Rotation group bias (time-in-sample eff ect) is the bias that occurs in a longitudinal survey when respondents provide less information as the number of times they are interviewed increases. Another kind of bias comes from the fact that respondents tend to remember serious events, like crimes, as having happened more recently than they actually did. At the fi rst interviews, there is no previous record in the survey of crimes, and, as a result, more crimes are reported in the fi rst interviews. Before 2006, the information from the first interviews in the NCVS was only used for bounding purposes and not in published crime rate estimates. However, in 2006, the Bureau of Justice Statistics (BJS) and the Census Bureau started to use the unbounded survey data for computing yearly crime rate estimates. As a result of the change in 2006, the estimated yearly crime rates have to be adjusted to eliminate the bias caused by lack of bounding information in the first interviews. We created a longitudinal property crime data fi le to track 26759 households through all seven of their NCVS interviews over a period of six years, recording the number of property crimes that were reported by each household in each interview. The main goal of this study is to develop a statistical model that will allow us to adjust the estimated crime rates for both time-in-sample and bounding biases. We fi t zero-infated count models, which allow extra zeros in standard Poisson count models, to model this property crime count data using a Bayesian approach. We also fit multinomial models to small domains longitudinal count data to study the bounding and time-in-sample e ffects. In the multinomial count model we also provide a way to deal with the missingness in the data. Adjustments to the estimated yearly household property crime rates are made using the model-fi tting results from the multinomial count models.
Elizabeth Stasny (Advisor)
Asuman Turkmen (Advisor)
Eloise Kaizar (Committee Member)
Catherine Calder (Committee Member)
150 p.
Statistics
rotation group bias, zero-inflated count models, multinomial count models, Bayesian models

Recommended Citations

Citations

  • Yang, H. (2016). Adjusting for Bounding and Time-in-Sample E ects in the National Crime Victimization Survey (NCVS) Property Crime Rate Estimation [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1452167047

    APA Style (7th edition)

  • Yang, Hui. Adjusting for Bounding and Time-in-Sample E ects in the National Crime Victimization Survey (NCVS) Property Crime Rate Estimation. 2016. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1452167047.

    MLA Style (8th edition)

  • Yang, Hui. "Adjusting for Bounding and Time-in-Sample E ects in the National Crime Victimization Survey (NCVS) Property Crime Rate Estimation." Doctoral dissertation, Ohio State University, 2016. http://rave.ohiolink.edu/etdc/view?acc_num=osu1452167047

    Chicago Manual of Style (17th edition)

osu1452167047
577
© 2016, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.