Skip to Main Content
 

Global Search Box

 
 
 

ETD Abstract Container

Abstract Header

Properties of Hurdle Negative Binomial Models for Zero-Inflated and Overdispersed Count data

Bhaktha, Nivedita
http://rave.ohiolink.edu/etdc/view?acc_num=osu1543573678017356

Abstract Details

2018, Doctor of Philosophy, Ohio State University, Educational Studies.
While studying rare events or behaviors in education and social science research, it is usual to encounter zero-inflated and/or overdispersed count data. Modeling the total number of serious violent incidents reported to the police that occurred on school campus from the 2010 School Survey on Crime and Safety (SSOCS) represents one such case and is the motivation for this dissertation. Generally, when overdispersion relative to Poisson is observed, Negative Binomial (NB) models are used. If there is additional zero-inflation in the NB model, then Hurdle Negative Binomial (HNB) model is used among other alternative models. The HNB models are fairly popular in Econometrics, Ecology, and Public Health, but is nascent in educational research. This study has been undertaken to learn the mechanisms of HNB models and make them accessible to educational researchers. To this effect, the HNB distribution was studied and its correct variance formula was derived using moment generating function. Simulation study was conducted to understand the effects of zero-inflation, overdispersion, and sample size on the HNB and NB models. It was also seen that in practice the flexibility of HNB models in accommodating different sets of predictors for binary and count parts were rarely made use of. The simulation study was expanded to examine the consequences of using different sets of predictors on conditional mean and probability of zero for varying levels of zero-inflation, overdispersion, and sample size. Finally, a demonstration of applying HNB models were presented using the 2010 SSOCS data. In this demonstration, detailed graphical exploratory data analysis for categorical variables are presented. Model fit evaluation has also been discussed. Finally, issues faced in applying HNB models were discussed and recommendations were provided for educational and social science researchers.
Ann O'Connell (Advisor)
Robert Cudeck (Committee Member)
Eloise Kaizar (Committee Member)
Jessica Logan (Committee Member)
161 p.
Educational Sociology; Social Research; Statistics
Zero-Inflation, Overdispersion, Count data, Variance formula of Hurdle Negative Binomial Model, Simulation Study, 2010 SSOCS data

Recommended Citations

Citations

  • Bhaktha, N. (2018). Properties of Hurdle Negative Binomial Models for Zero-Inflated and Overdispersed Count data [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1543573678017356

    APA Style (7th edition)

  • Bhaktha, Nivedita. Properties of Hurdle Negative Binomial Models for Zero-Inflated and Overdispersed Count data. 2018. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1543573678017356.

    MLA Style (8th edition)

  • Bhaktha, Nivedita. "Properties of Hurdle Negative Binomial Models for Zero-Inflated and Overdispersed Count data." Doctoral dissertation, Ohio State University, 2018. http://rave.ohiolink.edu/etdc/view?acc_num=osu1543573678017356

    Chicago Manual of Style (17th edition)

osu1543573678017356
4,379
© 2018, some rights reserved.Creative Commons License Properties of Hurdle Negative Binomial Models for Zero-Inflated and Overdispersed Count data by Nivedita Bhaktha is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License. Based on a work at etd.ohiolink.edu.
This open access ETD is published by The Ohio State University and OhioLINK.