Statistics 530 Name:__________________________ Final Exam In-class Part (Total Points: 50) 1. Be calm. 2. Good luck! Write TRUE or FALSE for the following with reasons: ( 3 points each) a. Robust regression with Huber weights is a special case of weighted regression. b. The number of parameters is proportional to the number of predictors for a non-linear regression model. c. For Pairwise comparison of treatment means the Tukey’s procedure is always the exact method for controlling Family-wise error rate. d. Since we are interested in a two-sided hypothesis in any general linear model the F test under consideration should be two-sided. e. For a multiple regression model of form: Yi = xi1 + xi2 xi3 + i The Error degrees of freedom is (n - 3). f. If the concomitant (numerical) variable (covariate) is itself related to the treatment, one way Analysis of Covariance (ANCOVA) is still valid. g. Since Logistic regression is essentially a transformation of the data, we could use yi=ln(Pi/(1-Pi)) transform on the data and then perform Simple Linear regression for the data relating Yi=b0+b1Xi +I h. Measurement errors in Y are NOT as problematic as measurement errors in X. i. Consider the inverse prediction problem, i.e. predicting X given Y. Since our choice of X and Y is arbitrary, we can switch the X and Y and perform simple regression. j. The standardized residuals are always smaller than the ordinary residuals. Problems: (20 points) 1. You are interested in testing whether two simple regression lines are parallel. Regression line 1: y = 1 +1 x + Regression line 2: y = 2 +2 x + What is the model, your hypothesis of interest, and testing procedure? Discuss this briefly. Hint: you may want to use General Linear Testing techniques. (5 points) 2. Is the following non-linear regression equation intrinsically linearizable? Why or why not? (6 points) a. f(x, ) = log(x)+2x b. Indicate how you would find starting values for (a), if you were to use non-linear methods. 3. Mention 4 classic signs that would make you suspect that you may have a problem with multicollinearity. 4. You are given the following table on rotated factor scores on nine variables Var1-Var9 using Varimax rotation. Based on the Table which variables would you consider relate to which factors? Varimax Rotated Factor Pattern Factor1 Factor2 Factor3 Factor4 Var1 -0.80065 0.03997 0.03137 0.05297 Var2 0.62302 0.30694 -0.27930 0.01978 Var3 0.07459 0.83236 0.03234 0.31625 Var4 0.03707 0.91395 -0.01210 -0.03861 Var5 0.27700 0.17264 0.28842 0.63901 Var6 0.07319 0.17414 0.80168 -0.11119 Var7 -0.07329 -0.15822 0.74358 0.11082 Var8 0.80224 0.00936 0.20642 0.05008 Var9 -0.18573 0.06387 -0.17869 0.82574 Have a great summer – I enjoyed teaching you!