【摘要】 根据自然复随机化法对福建古田地区人工降水试验的统计功效进行数值试验,结果表明,随机交叉设计试验2—3年能以80％的检出概率检验出20—30％的增雨效果(α=0.05),区域控制设计则要试验4—5年才能检验出相应的增雨效果;区域回归随机试验中,不同的统计检验方法:双比分析、多个事件检验和双样本回归分析,其检验功效差别不大,一般不超过5％;分层统计在一定条件下能提高统计功效,如层状云降水,由于其自然变差小,区域相关性高,分层统计其功效显著提高,积状云降水的统计功效则没有改善。区域相关性是影响统计功效的重要因素。更多还原

【Abstract】 The statistical power of precipitation enhancement experiment in Gutian region is computed numerically with naive method. The results of calculation showed, that the cloud seeding effects of 20-30% can be detected with power of 80% in 2-3 years duration with cross-over randomized design (α = 0.05), and the duration of 4-5 years is necessary to detect corresponding increase of rain with targetcontrol randomized design. In the randomized regression experiment of area, the deviations of statistical power are less than 5% generally with various evaluation methods: double ratio analysis, multiple event test and two-sample regression analysis. The statistical powers for cloud type stratification are different, which is larger in stratiform category than cumuliform category owing to the strong areal correlation.更多还原