Guidelines for the analysis. HW3 Q2 ST512 Sum2 2011 1. Create a table that summarize each regression model considered Model Overall F test RSquared AdjRsq Significant Reg coeff. 2. Look at the residual plot for each model. Do the residuals look randomly distributed? Is ther signs of heterogeneous variances? 3. Decide the best fit. Best looking residuals, high R squared, high Adj Rsq. 4. After selecting the best model, we run again this model to get the predicted in data set feedp and plot the predicted and observed so that we can demonstrate the goodness of fit of the selected model. Note that since Cu has only two distinct values, we plot two curves, one for each level of Cu. What are the regression equations for these two curves? Note that they are parallel which means that the intercept for Cu=0 is different from Cu=400. These two equations should show that. 5. 5 Finally, let’s look at the residual plot for the best fitted model once more. Do you see anything out of order? What about a point far away from the rest? Is it an outlier? 6. What changes should you expect if we drop the supposed outlier? Compare the average for Cu=0 vs Cu=400, and then look at significance for X2 in the best fit? How much the means change if we drop the outlier? N Cu Obs N Mean Std Dev Minimum Maximum ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ 0 30 30 2.7373333 1.0687987 1.3000000 5.0600000 400 30 30 2.8193333 1.1028394 1.1400000 5.0900000 ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ Outlier is identified Obs Cu 52 0 Additive 100 Feed Eff Ratio 3.87 x1 x2 100 0 x3 x4 10000 1000000 x5 0 x6 x7 0 0 pred resid 4.76733 -0.89733 New calculation for means without outlier Analysis Variable : FeedEffRatio N Cu Obs N Mean Std Dev Minimum Maximum _________________________________________________________________________________________ 0 30 29 2.6982759 1.0657059 1.3000000 5.0600000 400 30 30 2.8193333 1.1028394 1.1400000 5.0900000 7. models to choose from The REG Procedure Model: x1_x3_x4 Dependent Variable: FeedEffRatio Number of Observations Read 60 Number of Observations Used 60 Analysis of Variance Source DF Sum of Squares Mean F Value Square Model 3 66.77523 22.25841 Error 56 1.72460 0.03080 Corrected Total 59 68.49983 722.76 Root MSE 0.17549 R-Square 0.9748 Dependent Mean 2.77833 Adj R-Sq Coeff Var 6.31635 Pr > F <.0001 0.9735 Parameter Estimates Variable DF Parameter Estimate Standard t Value Pr > |t| Error Intercept 1 1.45560 0.05438 26.77 <.0001 x1 1 0.05411 0.00528 10.24 <.0001 x3 1 -0.00115 0.00013124 -8.75 <.0001 x4 1 0.00000942 8.617338E-7 10.93 <.0001 The REG Procedure The REG Procedure The REG Procedure models to choose from The REG Procedure Model: x1_x3_x2 Dependent Variable: FeedEffRatio Number of Observations Read 60 Number of Observations Used 60 Analysis of Variance Source DF Sum of Squares Mean F Value Square Model 3 63.19591 21.06530 Error 56 5.30393 0.09471 Corrected Total 59 68.49983 Pr > F 222.41 <.0001 Root MSE 0.30775 R-Square 0.9226 Dependent Mean 2.77833 Adj R-Sq Coeff Var 0.9184 11.07696 Parameter Estimates Variable DF Parameter Estimate Standard t Value Pr > |t| Error Intercept 1 1.64068 0.09674 16.96 <.0001 x1 1 0.00248 0.00415 0.60 0.5518 x3 1 0.00026522 0.00003982 6.66 <.0001 x2 1 0.00020500 0.00019865 1.03 0.3065 The REG Procedure The REG Procedure The REG Procedure models to choose from The REG Procedure Model: x1_x3_x2_x5_x6 Dependent Variable: FeedEffRatio Number of Observations Read 60 Number of Observations Used 60 Analysis of Variance Source DF Sum of Squares Mean F Value Square Model 5 63.22065 12.64413 Error 54 5.27918 0.09776 Corrected Total 59 68.49983 129.34 Root MSE 0.31267 R-Square 0.9229 Dependent Mean 2.77833 Adj R-Sq Coeff Var Pr > F <.0001 0.9158 11.25387 Parameter Estimates Variable DF Parameter Estimate Standard t Value Pr > |t| Error Intercept 1 1.65857 0.12673 13.09 <.0001 x1 1 0.00267 0.00596 0.45 0.6565 x3 1 0.00025786 0.00005721 4.51 <.0001 x2 1 0.00011554 0.00044807 0.26 0.7975 x5 1 -9.11607E-7 0.00002107 -0.04 0.9657 x6 1 3.683036E-8 2.022777E-7 0.18 0.8562 The REG Procedure The REG Procedure The REG Procedure models to choose from The REG Procedure Model: x1_x3_x4_x2 Dependent Variable: FeedEffRatio Number of Observations Read 60 Number of Observations Used 60 Analysis of Variance Source DF Sum of Squares Mean F Value Square Model 4 66.87609 16.71902 Error 55 1.62374 0.02952 Corrected Total 59 68.49983 Pr > F 566.31 <.0001 Root MSE 0.17182 R-Square 0.9763 Dependent Mean 2.77833 Adj R-Sq Coeff Var 6.18434 0.9746 Parameter Estimates Variable DF Parameter Estimate Standard t Value Pr > |t| Error Intercept 1 1.41460 0.05768 24.52 <.0001 x1 1 0.05411 0.00517 10.46 <.0001 x3 1 -0.00115 0.00012850 -8.93 <.0001 x4 1 0.00000942 8.43723E-7 11.16 <.0001 x2 1 0.00020500 0.00011091 1.85 0.0699 The REG Procedure The REG Procedure The REG Procedure Selected model y= x1 x3 x4 x2 The REG Procedure Model: x1_x3_x4_x2 Dependent Variable: FeedEffRatio Number of Observations Read 60 Number of Observations Used 60 Analysis of Variance Source DF Sum of Squares Mean F Value Square Model 4 66.87609 16.71902 Error 55 1.62374 0.02952 Corrected Total 59 68.49983 Pr > F 566.31 <.0001 Root MSE 0.17182 R-Square 0.9763 Dependent Mean 2.77833 Adj R-Sq Coeff Var 6.18434 0.9746 Parameter Estimates Variable DF Parameter Estimate Standard t Value Pr > |t| Error Intercept 1 1.41460 0.05768 24.52 <.0001 x1 1 0.05411 0.00517 10.46 <.0001 x3 1 -0.00115 0.00012850 -8.93 <.0001 x4 1 0.00000942 8.43723E-7 11.16 <.0001 x2 1 0.00020500 0.00011091 1.85 0.0699 Selected model y= x1 x3 x4 x2 The REG Procedure Model: x1_x3_x4_x2 Dependent Variable: FeedEffRatio Selected model y= x1 x3 x4 x2 Ob s 52 C Additiv u e 0 100 FeedEffRat io x1 x 2 x3 x4 x 5 x 6 x 7 pred resid 3.87 10 0 0 1000 0 100000 0 0 0 0 4.7673 3 0.8973 3
© Copyright 2024 Paperzz