气候影响评价的两种统计学方法--论广州天气对死亡率的影响

通过最优二分割方法,找出杜会、经济或生物的某种指标(本文用广州死亡率)所敏感的气象变量和临界值,用达到临界值的样本建立指标和气象变量的回归方程,提出了最优二分割--逐步回归模式。还研究了一种最优化天气分型的应用。检验表明,评价天气对死亡率的影响,分割-回归模式比一般回归好,最优化天气分型又比分割-回归好。

TWO STATISTICAL PROCEDURES USED IN CLIMATIC ASSESSMENT--AN INVESTIGATION OF WEATHER/MORTALITY RELATIONSHIP IN GUANGZHOU

Optimum partition-Stepwise Regression (OPSR) is the first procedure studied. By means of optimum partition, the threshold of the most sensitive meteorological variable to the index indicating specific conditions in society, economy and living beings is determined. Human mortality in the city of Guangzhou serves as the index in the study. Stepwise regression is then used to produce a weather/mortality model, based on the sample which consists of days beyond the threshold. The second procedure is the Optimized Classification of Weather (OCOW).Seventeen meteorological variables are treated by principal component analysis. Daily scores of a few top principal components are then classified using cluster analysis.The number of weather types is determined from the greatest drop of the sum of squares among groups.There is good correpondency between death rates and weather types.F-tests prove that OPSR is better than common regression, and OCOW is better than OPSR in terms of climatic impact assessment.