CH09 linear regression and correlation Medical Statistics医学统计学 知识讲稿.pptVIP

102 Test procedures n=8，r=0.8818，prep formula（9-19）: Confidence interval of population correlation coefficient Because the sampling distribution of correlation coefficient is skewed when ≠0 （also in large sample），we should make some transformation for variables to comply with normal distribution before estimating the confidence interval of . Concrete procedures Example 9-7 Estimate the 95% CI of population correlation coefficient, according to the value of example 9-5 According to the formula (9-22), make the anti-transformation, we get the 95%CI of the population correlation coefficient （0.4678, 0.9971） between the urine creatinine contents and the ages. 4 coefficient of determination Represent the ratio between the sum of squares for regression and total sum of squares. the calculate formula: （9-23） R2 value can range from 0 to 1. The numerical size represents the contribution proportion, also represent the percents of the correlate relationship counting for the total variances . Formula (9-23) shows that when the SStotal was fixed ，the size of the sum of squares for regression determined the size of the absolute value of correlation coefficient r . The closer the sum of squares for regression to the total sum of square , the closer the absolute value of r to 1, indicated that the result of the correlation was better. Example (9-23),the linear correlation coefficient r=0.8818 between the ages and the urine creatinine contents of the eight children ,R2＝0.7775，show that the age can explain the 77.5% of the variances of the urine creatinine concentration, the other 22％ variances can’t explain by age. 5 Cautions about linear regression and correlation 1 According to the analytic purpose to select variables and statistic methods Linear correlation used to explain the strength and direction of the correlation between the two variables, x and y are the same important. Linear regression used to f