应用水文统计学复习.pptVIP

Normally, we use a dimensionless (scale free) measure to describe the association between the X and Y by dividing the covariance by the product of the individual standard deviations. Correlation coefficient, (Pearson’s rho) , is a measure of the strength of their linear association between X and Y. Sample correlation coefficient Let (x1, y1), (x2, y2), ……, (xn, yn) denote a random sample of n pairs of observations on the random variables X and Y. We have sample correlation coefficient, r, given below: Practically, Hypothesis of no linear association Test for Population Correlation Let r be the sample correlation coefficient, calculated from a random sample of n pairs of observations from a joint normal distribution.  3. Linear regression It is called the population regression line of Y on X (称为Y依X的理论回归直线），where are coefficients (constants) and is error term, a random variable with mean 0, i.e. is the slope of the regression line.  对 有以下假定： 独立性：对于任意i和j， 零均值性： 共方差性： 正态性： i.e. 是相互独立且服从同一正态分布 的随机变量。 Least squares estimation and Sample regression line (1) Let (x1, y1), (x2, y2), ……, (xn, yn) be a sample of n pairs of observations on a process with population regression line , the best: The best estimates can be obtained by minimizing the residuals via least square method, i.e. We have Least squares estimation and Sample regression line (2) It can be shown that the resulting estimates are: where are the respective means. The line is called the sample regression line of Y on X (Y 依 X 的回归直线）. Discussion on (1) 1. 互相独立的两个随机变量无相关关系；但 无相关关系的两个随机变量不一定独立。 （一个园） 由f(x,y)的对称性，ax = ay = 0。 于是，  Discussion on (2) 2. Correlation does not imply causation （cause-and-effect relationship）  Discussion on (3) 3.