The statistical implications of autocorrelation for detection in environmental health assessment

摘要: Many environmental health and risk assessment techniques models aim at estimating the fluctuations of selected biological endpoints through time domain as a means assessing changes in environment or probability particular measurement level occurring. In either case, estimates sample variance mean are crucial to making appropriate statistical inferences. The commonly employed for both measures presume data were generated by covariance stationary process. such cases, observations treated independently identically distributed classical testing methods applied. However, if assumption stationarity is violated, resulting biased. bias compromises procedures increasing detecting significance tests differences. This can lead inappropriate decisions being made about severity damage. Accordingly, it argued that sets be examined correlation adjustments required estimators before they used hypothesis testing. Only then credible scientifically defensible decision makers regulators.