Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Homework # 3 Key Chapter 4: 12, 15, 16, plus a fourth question not from the book 80 total points 4.12 Predicted therapy satisfaction = 1 + (.4)*depression score (2pts each) a) 1 + .4(5) = 3 b) 1 + .4(6) = 3.4 c) 1 + .4(7) = 3.8 d) 1 + .4(8) = 4.2 e) 1 + .4(9) = 4.6 f) 1 + .4(10) = 5 4.15 They may do this one with z-scores or using raw scores. (This is also the data from one of the problems on the last homework, so they may have “reused” some of their old computations (r = -0.985). The problem also gives some information (SStotal = 84) X Y 1 1 2 4 Sum 8 Mean 2 SS 6 10 8 4 -2 20 5 84 stdev 4.58 1.22 Using Z-scores ZX ZY ZX ZY -0.82 -0.82 0.0 1.64 1.09 0.66 -0.22 -1.53 -0.89 -0.54 0.0 -2.51 -3.94 r = -0.985 (X X) -1 -1 0 2 Using raw scores (Y Y ) (X X) (Y Y ) 5 -5 3 -3 -1 0 -7 -14 SP = -22 SP 22 r SSX SSY 6 * 84 r = -0.985 (a) Determine the raw-score prediction formula for predicting anxiety from dexterity (10 pts) Standardized Beta = r = -0.985 SP 22 1 slope 3.67 SSX 6 ZYˆ (0.985)Z X intercept Y X 0 computing intercept (what does Y = when X = 0?) (X X) (0 2) ZX 0 1.64 sX 1.22 ZYˆ (0.985)(1.64) 1.615 Y sY ZY) Y YX 0 4.58 1.615 5 12.3 Computing the raw score for slope 1 5.0 3.67 2.0 12.34 Yˆ 12.34 (3.67)X (X X) (1 2) 0.82 sX 1.22 ZYˆ (0.985)(0.82) 0.81 Z X 1 Y sY ZY) Y YX 1 4.58 0.81 5 8.71 ) ) slope YX 1 YX 0 8.71 12.3 3.6 Yˆ 12.3 (3.6)X b) (specific numbers will vary with slight round off differences, give credit if they’re in the right ballpark) (8 pts) Y 12.3 3.67 1 8.63 ) (Y Y ) 10 8.63 1.37 ) (Y Y )2 1.8 8 Y 12.3 3.67 1 8.63 8 8.63 0.63 0.4 2 4 Y 12.3 3.67 2 4.96 4 4.96 0.96 0.9 4 -2 2 2.38 0.38 0.14 X Y 1 10 1 Y 0 1 X Y 12.3 3.67 4 2.38 SSerror = 3.2 c) plot the scatterplot with the regression line (4 pts) 10 8 6 4 2 1 2 3 4 5 -2 d) figure the error and squared error for each of the four predictions (see table in b) (4pts) e) find the proportionate reduction in error using SSerror and SStotal. (4 pts) r2 SStotal SSerror 84 3.2 0.96 SStotal 84 f) take the square root of (e) and compare it to the correlation coefficient . (4 pts) r 2 0.96 0.98 They are essentially the same (except you lose the direction information because squaring a negative number leads to a positive number [and square rooting a positive leads to a positive]). To regain the direction of the relationship you need to examine the sign of your slope g) answers will vary here. They should try to explain what regression is good for (making predictions) and how it works. . (4 pts) 4.16 Same data as in 4.15, but now they switch the X and Y variables (predict dexterity from anxiety). They may do this one with z-scores or using raw scores. (This is also the data from one of the problems on the last homework, so they may have “reused” some of their old computations (r = -0.985). The problem also gives some information (SStotal = 6) X Y 10 8 4 -2 Sum 20 Mean 5 SS 84 1 1 2 4 8 2 6 stdev 1.22 4.58 Using Z-scores ZX ZY ZX ZY 1.09 0.66 -0.22 -1.53 -0.82 -0.82 0.0 1.64 -0.89 -0.54 0.0 -2.51 -3.94 r = -0.985 (X X) 5 3 -1 -7 Using raw scores (Y Y ) (X X) (Y Y ) -1 -5 -1 -3 0 0 2 -14 SP = -22 SP 22 r SSX SSY 6 * 84 r = -0.985 (a) Determine the raw-score prediction formula for predicting anxiety from dexterity (12pts) Standardized Beta = r = -0.985 SP 22 1 slope 0.26 SSX 84 ZYˆ (0.985)Z X intercept Y X 0 computing intercept (what does Y = when X = 0?) (X X) (0 5) ZX 0 1.09 sX 4.58 ZYˆ (0.985)(1.09) 1.075 Y sY ZY) Y YX 0 1.22 1.075 2 3.31 1 2.0 0.26 5.0 3.31 Yˆ 3.31 (0.26)X Computing the raw score for slope (X X) (1 5) Z X 1 0.87 sX 4.58 ZYˆ (0.985)(0.87) 0.86 Y sY ZY) Y YX 1 1.22 0.86 2 3.05 ) ) slope YX 1 YX 0 3.05 3.31 0.26 Yˆ 3.31 (0.26)X b) (specific numbers will vary with slight round off differences, give credit if they’re in the right ballpark) (8pts) Y 3.31 0.26 10 0.71 ) (Y Y ) 1 0.71 0.29 ) (Y Y )2 0.08 1 Y 3.31 0.26 8 1.23 1 1.23 0.23 0.05 4 2 Y 3.31 0.26 4 2.27 2 2.27 0.27 0.07 -2 4 Y 3.31 0.26 2 3.83 4 3.83 0.17 0.03 X Y 10 1 8 Y 0 1 X SSerror = 0.23 c) plot the scatterplot with the regression line (4pts) 5 4 3 2 1 -2 2 4 6 8 10 d) figure the error and squared error for each of the four predictions (see table in b) (4pts) e) find the proportionate reduction in error using SSerror and SStotal. (4pts) r2 SStotal SSerror 6 0.23 0.96 SStotal 6 f) take the square root of (e) and compare it to the correlation coefficient (4pts) r 2 0.96 0.98 They are essentially the same (except you lose the direction information because squaring a negative number leads to a positive number [and square rooting a positive leads to a positive]). To regain the direction of the relationship you need to examine the sign of your slope g) answers will vary here. They should try to explain what regression is good for (making predictions) and how it works. (4pts) Fourth question) Additionally answer the following questions based on the data contained in height.sav. Note: This is an SPSS datafile. You'll need to use SPSS to do these questions (so don't try to open the file on a computer that doesn't have SPSS on it. It won't work.) Compute the following regression models predicting HEIGHT: (a) predict height with average height of parents (b) predict height with average calcium intake (c) predict height with weight (d) predict height with average height of parents and weight (e) predict height with average height of parents and average calcium intake (f) predict height with all three variables (avg height of parents, avg calium, weight) Based on the SPSS output (you don't need to print it out and include it), report the r2 for each model. Which model do you think is the "best?" Explain why. (12 pts) Model 1 2 3 4 5 6 Variables average height of parents average calcium intake weight average height of parents weight average height of parents average calcium intake average height of parents average calcium intake weight Unstandardized ß 0.8 0.41 0.79 0.49 0.47 0.81 -0.01 0.47 0.03 0.48 Model r2 0.641 0.168 0.631 0.767 0.621 0.768 Model 6 accounts for slightly more variance than model 4. However, model 4 has fewer variables in it. Since model 4 is has fewer variables and accounts for nearly the same amount of variance (as model 6) that should