Applications of Bayesian Statistics to Occupational Exposure Assessment

While working as a graduate student at the University of Minnesota, I became involved in the Gulf Long-term Follow-up Study (GuLF STUDY). During the course of my work on this project (which culminated in Quick et al., 2014), I was introduced to Drs. Tran Huynh and Gurumurthy Ramachandran and their research related to occupational exposure assessment in the case of censored (or otherwise limited) data. Shortly thereafter, we began investigating how existing Bayesian statistical approaches could be used to leverage prior information to yield improved estimates of occupational exposures.

Published work:

  • Groth, C.P., Huynh, T., Banerjee, S., Ramachandran, G., Stewart, P.A., Quick, H., Sandler, D.P., Blair, A., Engel, L.S., Kwok, R.K., and Stenzel, M.R. (2021). "Linear relationships between total hydrocarbons and benzene, ethylbenzene, toluene, xylene, and hexane during the Deepwater Horizon response and clean-up." Published online ahead of print. Ann. Work Exp. Health. [annweh]

  • Huynh, T.B., Groth, C.P., Ramachandran, G., Banerjee, S., Stenzel, M., Quick, H., Blair, A., Engel, L.S., Kwok, R.K., Sandler, D.P., and Stewart, P.A. (2020). "Estimates of occupational inhalation exposures to six oil-related compounds on the four rig vessels responding to the Deepwater Horizon oil spill." Ann. Work Exp. Health. Published online ahead of print. [annweh]

  • Huynh, T., Ramachandran, G., Quick, H., Hwang, J., Raynor, P.C., Alexander, B.H., and Mandel, J.H. (2019). "Ambient fine aerosol concentrations in multiple metrics in taconite mining operations." Ann. Work Exp. Health, 63, 77-90. [annweh]

  • Quick, H., Huynh, T., and Ramachandran, G. (2017). “A method for constructing informative priors for Bayesian modeling of occupational hygiene data.” The Annals of Work Exposures and Health, 61, 67-75. [annweh]

    • Describes how one could construct informative priors for occupational hygiene data based on the concept of "past data". In particular, the goal of this work is to specify priors with a predetermined, user-specified "prior sample size", which can be directly compared to the sample size of the data. Specifying priors in this form requires estimates of the geometric mean (GM) and the geometric standard deviation (GSD), in addition to prior sample size, n0. We also discuss the option to restrict the parameter space -- e.g., restricting the 95th percentile to be less than a user-specified threshold (e.g., twice the occupational exposure limit) when the objective is to determine which AIHA exposure category the given exposure group belongs to.

    • R code and WinBUGS code corresponding to the illustrative example in Section 4 (Last updated 09/16/2016)

  • Huynh, T., Quick, H., Ramachandran, G., Banerjee, S., Stenzel, M., Blair, A., Sandler, D., Engel, L., Kwok, R.K., and Stewart, P.A. (2016). “A comparison of the β-substitution method and a Bayesian approach for analyzing left-censored data for the GuLF STUDY.” The Annals of Occupational Hygiene, 60, 56-73. [annhyg]

    • Compares the popular β-substitution approach of Ganser and Hewett (2010) to a Bayesian approach for analyzing left-censored data.

  • Quick, H., Groth, C., Banerjee, S., Carlin, B.P., Stenzel, M.R., Stewart, P.A., Sandler, D.P., Engel, L.S., and Kwok, R.K. (2014). “Exploration of the use of Bayesian modeling of gradients for censored spatiotemporal data from the Deepwater Horizon oil spill.” Spatial Statistics, 9, 166-179. [spasta]

      • Extends the work of Quick et al. (2015) to estimate the spatiotemporal gradient process in the presence of censored data.