The impact of model untercertainty on biological parameter estimates. (English) Zbl 1068.62108

Burr, Tom; Gattiker, James

Far East J. Theor. Stat. 13, No. 2, 153-174 (2004).

Summary: Classical and molecular epidemiology rely strongly on model-based estimates of biological parameters. For example, classical epidemiology models lead to estimates of the per-sex-act or per-dirty-needle-use probability of HIV infection. Molecular epidemiology models of the evolution of HIV at the DNA level lead to estimates of the time to the most recent common ancestor of present-day HTV cases, or of population dynamics such as the rate of growth of the epidemic. Because biological models play a central role in these and many other estimates, as well as in our overall understanding of disease etiology, it is important to understand the sensitivity of conclusions to model uncertainty.
We provide a brief overview of the impact of model uncertainty, describe what is known about its impact in a few examples, and present a detailed example involving subdivided populations in the estimated per-dirty-needle use probability of HTV infection.

MSC:

62P10	Applications of statistics to biology and medical sciences; meta analysis
92D30	Epidemiology
62N02	Estimation in survival analysis and censored data
62F10	Point estimation

Keywords:

Molecular epidemiology; model uncertainty