Effect of sampling uncertainty and spatial correlation on an empirical Bayes procedure for combining site and regional information |
| |
Authors: | George Kuczera |
| |
Institution: | Department of Civil Engineering, Monash University, Clayton, Vic. 3168 Australia |
| |
Abstract: | The effect of sampling uncertainty and spatial correlation on the pooling of site and regional information is studied in the context of the empirical Bayes (EB) normal probability model chosen because of its simplicity and generality. Because the EB model parameters must be evaluated from observed hydrologic data they are subject to uncertainty which can be considerable when samples are small and inter-site correlation exists. Procedures are developed which permit approximate assessment of the effect on inference of uncertainty in these EB parameters. Hence, the gains that pooled inference can make over site or regional inference can be realistically evaluated. It is also shown that when the site estimate differs much more than expected from the regional estimate pooling of information can be counterproductive in the sense of site or regional inference being more precise. |
| |
Keywords: | |
本文献已被 ScienceDirect 等数据库收录! |
|