Survey
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
1 The slope 1 is the true average change in the dependent variable y associated with a 1-unit increase in x. The slope of the least squares line b1 ˆ1 gives an estimate of the true slope This estimate depends on the values of Y, which are random. If we can determine the sampling distribution of the estimate, we can perform inference for the true slope. 2 The estimator of the slope is b1 ˆ1 x x Y Y S S x x i i 2 i where xy ci xi x / S xx ci Yi Y ciYi xx . This is a linear combination of normal random variables Yi , and thus it has a normal distribution. 3 E ciYi ci E Yi ci 0 1 xi 0 ci 1 ci xi 0 1 ci xi x 1 V ciY c V Yi 2 i xi x S 2 xx 2 2 2 S xx 4 Recall that in simple linear regression we estimate 2 using SSE yi2 ˆ0 yi ˆ1 xi yi n2 2 s (which the book calls ). n2 Then the estimated standard deviation of the s ˆ estimator 1 is sˆ . 1 S xx 5 The assumptions of the simple linear regression model then imply that the standardized variable T ˆ1 1 S ˆ 1 ˆ1 1 S / S xx has a t distribution with n-2 d.f. 6 1 Confidence intervals and hypothesis tests for are then carried out in the usual manner. A 100 1 % confidence interval for 1 is ˆ1 t /2,n2 sˆ 1 7 Null hypothesis: H 0 : 1 10 Test statistic: t ˆ1 10 / Sˆ 1 Alternative hypothesis H a : 1 10 H a : 1 10 H a : 1 10 Rejection region t t ,n2 t t ,n2 t t /2,n2 or t t /2,n2 The test H 0 : 1 0 versus H a : 1 0 usefulness of the model. tests the 8 The sample correlation coefficient r gives a measure of the linear relationship among X andY. Whereas for linear regression the X variable is fixed, here it doesn’t matter which variable is called X, and which is calledY. The statistic r is related to the coefficient of determination in simple linear regression, and forms an estimate of the population correlation coefficient . 9 The sample correlation coefficient for the n pairs x1 , y1 , , xn , yn is r x x y y x x y y i 2 i i i 2 S xy S xx S yy Recall that b1 S xy / S xx , so that the estimated slope and r have the same sign. 10 The value of r is independent of the units in which x and y are measured r lies in the interval 1 r 1 r = 1 if and only if all xi , yi pairs lie on a straight line with positive slope, and r = -1 if and only if all pairs lie on a straight line with negative slope. The square of the correlation coefficient gives the value of the coefficient of determination from fitting the simple linear regression model. r measures the degree of the linear relationship 11 Weak .5 r .5 Moderate .8 r .5 or .5 r .8 Strong r .8 or r .8 The rationale for calling correlations weak even when they are as large in absolute value as .5 is 2 r .25 , so that if r .5 the that even in that case linear model explains at most 25% of the observed variation, which is not very impressive. 12 We can think of the pairs xi , yi as having been drawn from a bivariate population of pairs, with some joint pmf or pdf, and correlation . When the joint pdf is bivariate normal, one can carry out inference for . Let (X,Y) be bivariate normal with respective 2 2 , , means 1 2 , variances 1 2 , and correlation coefficient . 13 If X = x, it can be shown that the (conditional) distribution ofY is normal with mean Y x 2 1 2 / 1 2 x / 1 and variance 1 2 22 This fits the simple linear regression model with 0 2 1 2 / 1 , 1 2 / 1 , and 2 1 2 22 . 14 The implication is that if the observed pairs are actually drawn from a bivariate normal distribution, then the simple linear regression model is an appropriate way of studying the behavior ofY given X=x. If 0 , then Y x 2 , independent of x. 15 When H 0 : 0 is true, the test statistic T R n 2 / 1 R 2 has a t distribution with n-2 d.f. Alternative hypothesis Ha : 0 Ha : 0 Ha : 0 Rejection region t t ,n2 t t ,n2 t t /2,n2 or t t /2,n2 16 The null hypothesis 0 states that there is no linear relationship between X and Y in the population. In the context of regression analysis, we used ˆ1 / sˆ to test for the absence of a linear relationship (1 0 ). Since r n 2 / 1 r 2 ˆ1 / sˆ the tests are equivalent. 1 1 17