Varition and Prediction Intervals

10-4 Variation and Prediction Intervals Explained and unexplained variation In this section, we study two measures used in correlation and regression studies. (The coefficient of determination and the standard error of estimate.) We also learn how to construct a prediction interval for y using a regression line and a given value of x. To study these concepts, we need to understand and calculate the total variation, explained deviation, and the unexplained deviation for each ordered pair in a data set. Assume that we have a collection of paired data containing the sample point (x , y), that 𝑦 is the predicted value of y, and that the mean of the sample y-values is 𝑦. The total variation about a regression line is the sum of the squares of the differences between the y-value of each ordered pair and the mean of y. total variation = (𝒚 − 𝒚)𝟐 The explained variation is the sum of the squared of the differences between each predicted y-value and the mean of y. explained variation = (𝒚 − 𝒚)𝟐 The unexplained variation is the sum of the squared of the differences between the y-value of each ordered pair and each corresponding predicted y-value. unexplained variation = (𝒚 − 𝒚)𝟐 The sum of the explained and unexplained variations is equal to the total variation. Total variation = Explained variation + Unexplained variation As its name implies, the explained variation can be explained by the relationship between x and y. The unexplained variation cannot be explained by the relationship between x and y and is due to chance or other variables. Consider the advertising and sales data used throughout this section with a regression line of 𝑦 = 50.729 x + 104.061. Using the data point (2.0, 220) we can find the total, explained, and unexplained variation: The Coefficient of determination The coefficient of determination r2 is the ratio of the explained variation to the total variation. 𝑟2 = 𝑒𝑥𝑝𝑙𝑎𝑖𝑒𝑛𝑑 𝑣𝑎𝑟𝑖𝑎𝑡𝑖𝑜𝑛 𝑡𝑜𝑡𝑎𝑙 𝑣𝑎𝑟𝑖𝑎𝑡𝑖𝑜𝑛 We can compute 𝑟 2 by using the definition or by squaring the linear correlation coefficient r. Ex 1) The correlation coefficient for the following advertising expenses and company sales data is 0.913. Find the coefficient of determination. What does this tell you about the explained variation of the data about the regression line? About the unexplained variation? (r= 0.913 suggests a strong positive linear correlation) 𝒓𝟐 = 0.834 About 83.4% of the variation in the company sales can be explained by the variation in the advertising expenditures. About 16.6% of the variation is unexplained and is due to chance or other variables. xy x2 225 540 5.76 50,625 1.6 184 294.4 2.56 33,856 2.0 220 440 4 48,400 2.6 240 624 6.76 57,600 1.4 180 252 1.96 32,400 1.6 184 294.4 2.56 33,856 2.0 186 372 4 34,596 2.2 215 473 4.84 46,225 15.8 1634 3289.8 32.44 337,558 𝑦 =(1634/8) =204.25 𝑥 =(15.8/8)=1.975 , Advertising expenses Company sales (1000s of $), x (1000s of $), y 2.4 y2 Sums The Standard Error of Estimate The Standard Error of Estimate se is the standard deviation of the observed y-values about the predicted 𝑦-value for a given x-value. It is given by 𝑠𝑒 = Or as the following equivalent formula: (𝑦 − 𝑦 )2 𝑛−2 𝑠𝑒 = 𝑦 2 − 𝑏0 𝑦 − 𝑏1 𝑛−2 𝑥𝑦 Ex 2) The regression equation of the advertising expenses and company sales data in example 1) is 𝑦 = 50.729 x + 104.061 Find the standard error of estimate. 𝑦 (𝑦 − 𝑦)2 x y 2.4 225 225.81 0.6561 1.6 184 185.23 1.5129 2.0 220 205.52 209.6704 2.6 240 235.96 16.3216 1.4 180 175.08 24.2064 1.6 184 185.23 1.5129 2.0 186 205.52 381.0304 2.2 215 215.66 0.4356 Sum 635.3463 The standard error of estimate of the company sales for a specific advertising expense is about $10,290. In chapter 7, we saw that point estimates will not give us any information about how accurate they might be. Thus, we developed confidence interval estimates to overcome this advantage. In this section we follow the same approach to construct a prediction interval. A prediction interval is an interval estimate of a predicted value of y. Given a linear regression equation 𝑦 = 𝑏0 + 𝑏1 𝑥 and x0, a specific value of x, a prediction interval for y is 𝑦−𝐸 <𝑦 <𝑦+𝐸 Where 𝐸= 𝑡𝛼 𝑠𝑒 2 1 𝑛 𝑥0 − 𝑥 2 1+ + 𝑛 𝑛 𝑥2 − 𝑥 2 With n-2 degrees of freedom. Ex3) Using the results of previous example, construct a 95% prediction interval for the company sales when the advertising expenses are $2100. What can you conclude?