Files and links (1)
CC BY-NC-SA V4.0, Open Access
Dissertation
Analyzing workplace accident underreporting: a systematic review, a Monte Carlo simulation, and a real data application
Washington State University
Doctor of Philosophy (PhD), Washington State University
2023
DOI:
https://doi.org/10.7273/000005095
Abstract
This dissertation is made up of three studies considering the most appropriate measurement and analysis of workplace accident underreporting. A systematic review (Study 1) highlighted a heterogeneous set of choices in terms of measurement and analytical strategies. Multi-item recognition-based scales seem to perform better compared to recall measures. Likewise, scholars used several analytical strategies when estimating count models. However, some of those might be inappropriate as they might lead to biased estimates. The Monte Carlo simulation (Study 2) showed that General Linear Model (GLM) with a negative binomial and Poisson distributions and Linear Model (LM) with raw outcomes returned similar parameter estimates (i.e., average marginal effects) in large samples, whereas the LM with log-transformed outcomes and GLM with binary outcomes yielded biased estimates. Analyses on a real dataset (Study 3) added two pieces of evidence: standard errors and model fit. LM with raw outcomes, GLM with a Poisson distribution, and GLM with a negative binomial distribution returned comparable standard errors in large samples, but standard errors in the LM with log-transformed outcomes and GLM with binary outcomes were underestimated. Compared with the GLM with negative binomial, LM with raw outcomes and GLM with a Poisson distribution showed worse fit to the data, likely capturing severe violations of distributional assumptions. Overall, these results suggest that LM with raw outcomes, GLM with negative binomial and Poisson distributions are equivalent in medium samples (N > 500). Six best practice recommendations are also discussed.
Metrics
5 File views/ downloads
21 Record Views
Details
- Title
- Analyzing workplace accident underreporting
- Creators
- Andrea Bazzoli
- Contributors
- Tahira M. Probst (Advisor)Jeremy M. Beus (Committee Member)Elizabeth A. Canning (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Psychology, Department of
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Publisher
- Washington State University
- Number of pages
- 88
- Identifiers
- 99901019536201842
- Language
- English
- Resource Type
- Dissertation