Page 435 - Contributed Paper Session (CPS) - Volume 2
P. 435
CPS1916 Erica P. et al.
This is certainly a major advantage of using Bayesian hierarchical models,
which provide a general adaptable way to formulate a broad range of models
and structures, as in our case measurement error or dependency structures,
and more generally any additional random effect which might be needed in
the analysis. Moreover, the use of a Bayesian framework allows to incorporate
prior knowledge in the analysis, for example about the error component and
parameters, as well as to reflect the prior uncertainty in the posterior
distributions of parameters of interest. Of course, this requires some
knowledge about the error component, in order to properly formulate the
measurement error level and to assign reasonable priors. Note that this
requirement is not specific to the Bayesian framework, but rather to any error
modelling strategies. In fact, it is always necessary to know the error structure
(i.e. classical measurement error and its distribution), as well as the error
variance, in order to formulate an identifiable error model (Gustafson, 2005).
In practice, it is not always straightforward to obtain such information, and
often assumptions about the error distribution and parameters are vague or
potentially incorrect. The advantage of Bayesian models is that the uncertainty
in such assumptions can easily be accounted for and propagated to posterior
distributions of corrected estimates.
References
1. Carroll, R.J., D. Ruppert, L.A. Stefanski, and C.M. Crainiceanu. 2006.
Measurement Error in Nonlinear Models, a Modern Perspective. Boca
Raton: Chapman; Hall.
2. Dominici, F., S. Zeger, and J. Samet (2000). “A measurement error model
for time- series studies of air pollution and mortality.” Biostatistics 1 (2),
157–175
3. Edwards, J.K., and A.P. Keil. 2017. “Measurement Error and Environmental
Epidemiology: A Policy Perspective.” Current Environmental Health
Reports 4 (1): 79–88.
4. Goldman, G.T., J.A. Mulholland, A. G. Russell, M.J. Strickland, M. Klein, L.A.
Waller, and P.E. Tolbert. 2011. “Impact of Exposure Measurement Error in
Air Pollution Epidemiology: Effect of Error Type in Time-Series Studies.”
Environmental Health 10 (61).
5. Gustafson, P. 2005. “On Model Expansion, Model Contraction,
Identifiability and Prior Information: Two Illustrative Scenarios Involving
Mismeasured Variables.” Statistical Science 20: 111–40.
6. Muff, S., M. Ott, J. Braun, and L. Held (2017). “Bayesian two-component
measurement error modelling for survival analysis using INLA - A case
study on cardiovascular disease mortality in Switzerland.” Computational
Statistics and Data Analysis 113, 177–193.
424 | I S I W S C 2 0 1 9