STA 6167 – Exam 3 – Spring 2012 PRINT Name __________________________________________________________ Conduct all tests at = 0.05 significance level. Q.1. A study was conducted to relate incidence of West Nile virus in horses among the counties of South Carolina. The dependent variable was the number of cases of West Nile virus in horses, the exposure variable was the number of farms (t), and the predictor (independent) variables were PBR (# Bird Cases of West Nile / Human Population) and Human Density (HD), and the interaction between PBR and HD. Three Poisson Regression models are fit: Model 0: ln(/t) = ln() = ln(t) + Model 1: ln(/t) = PBR + HD ln() = ln(t) + PBR + HD Model 2: ln(/t) = PBR + HD + PBR*HD ln() = ln(t) + PBR + HD + PBR*HD Note: In the following tables, Residual Deviance(ModelA)-Residual Deviance(ModelB) = -2(logLA – logLB) where model B has more parameters (predictor variables) than model A. Parameter Beta0 Beta1 Beta2 Beta3 Res.Dev. Res. Df Model0 Estimate StdErr -5.92 0.14 #N/A #N/A #N/A #N/A #N/A #N/A 109.32 45 Model1 Estimate StdErr -7.18 0.35 7492 1094 0.00260 0.00127 #N/A #N/A 66.32 43 Model2 Estimate StdErr -6.98 0.37 5330 1596 0.00059 0.00167 26.00 12.38 62.01 42 p.1.a. Give the (estimated) predicted values for Dorchester County, where t = 314 farms, PBR = 0.00022, and HD = 168. p.1.a.i. Model 0: ln() = ______________________________________________ ___________________ p.1.a.ii. Model 1: ln() = ______________________________________________ ___________________ p.1.a.iii. Model 2: ln() = ______________________________________________ ___________________ p.1.b. Use the likelihood ratio test to test whether the main effects of PBR and/or HD are significant, by comparing model 1 and model 0: H0: versus HA: and/or≠ 0 Test Statistic _____________________________________ Reject H0 if the test statistic is larger/smaller than ________ p.1.c. Use the Wald test to test whether PBR and HD are significant effects, controlling for the other predictor: p.1.c.i. H0:PBR Not associated with West Nile Incidence in horses, controlling for HD: Test Statistic ______________________________________ Reject H0 if the test statistic is larger/smaller than ________ p.1.c.ii. H0:HD Not associated with West Nile Incidence in horses, controlling for PBR: Test Statistic ______________________________________ Reject H0 if the test statistic is larger/smaller than ________ p.1.d. Test whether the PBR/HD interaction is significant (controlling for main effects): LR Test Stat ____________ Wald Test Stat ____________ Reject H0 if the test statistic is larger/smaller than _________ Q.2. A study based on a random sample of 50 horror movies, identified 465 characters, and classified them in terms of if they were Female (F, 1=Yes,0=No), whether they had Sexual Activity (SA, 1=Yes,0=No), and whether they Lived (L) through the film. The dependent variable was whether the character Lived through the film (1=Yes, 0=No). The following sequence of models were fit with respect to logit(P(L=1)) = ln(P(L=1)/(1-P(L=1))): Model 0: Model 1: F Model 2: F + SA Model 3: F + SA + F*SA Variables in the Equation Va riables in the Equa tion St ep 0 Constant B -1. 549 S. E. .119 W ald 168.165 df 1 Sig. .000 Step a 1 Ex p(B) .213 Female Constant Step a 1 Female SexAct Constant S.E. .246 .289 .190 Wald 8.411 7.087 80.311 S.E. .242 .182 Wald 6.937 106.558 df 1 1 Sig. .008 .000 Exp(B) 1.894 .154 a. Variable(s) entered on s tep 1: Female. Variables in the Equation Variables in the Equation B .713 -.770 -1.699 B .639 -1.874 df 1 1 1 Sig. .004 .008 .000 a. Variable(s) entered on s tep 1: SexAct. Exp(B) 2.040 .463 .183 Step a 1 Female SexAct FmSA Constant B .808 -.510 -.428 -1.749 S.E. .279 .447 .583 .205 Wald 8.408 1.300 .537 72.980 df 1 1 1 1 Sig. .004 .254 .463 .000 Exp(B) 2.243 .601 .652 .174 a. Variable(s) entered on s tep 1: FmSA. P.2.a. Based on the Wald tests, which statement best describes the results (mark the best answer, making use of the order of models): p.2.a.i. Neither F, SA, or their interaction significant p.2.a.iii. Only F is significant p.2.a.ii. Interaction is significant, neither of the main effects are significant p.2.a.iv. Only SA is significant p.2.a.v. F and SA significant, their interaction is not. P.2.b. Based on Model 2, give the predicted probabilities of L=1 for each of the following groups: p.2.b.i. Female/Sexually Active: p.2.b.ii. Male/Sexually Active: p.2.b.iii. Female/Not Sexually Active: p.2.b.iv. Male/Not Sexually Active: p.2.c. Based on your results from p.2.b., show the interpretation of exp(B2)=0.463 for model 2 p.2.c.i. Males: __________________________________ / ________________________________ = 0.463 p.2.c.ii. Females: __________________________________ / ________________________________ = 0.463 Q.3. A nonlinear regression model is fit, relating Y (responsiveness, as a percentage of maximum possible responsiveness) to X (Dose) for 19 individuals. The model fit is: 1 Y 0 1 1 1 X 2 The estimates and standard errors are given below. Parameter Estimates Parameter b0 b1 b2 Es timate 100.340 6.480 4.816 Std. Error 1.174 .194 .028 95% Confidence Interval Lower Bound Upper Bound 97.851 102.829 6.068 6.892 4.756 4.875 p.3.a. Give the fitted values for the following X-values: p.3.a.i. X=0 p.3.a.ii. X=1 p.3.a.iii. X=8 p.3.a.iv. X=9 p.3.b. At what dose does the curve reach halfway to its maximum responsiveness? 120 100 e vs Y-hat 80 3 2 Y 40 residual Y-hat 60 1 0 -1 20 0 20 40 60 80 100 120 -2 0 -3 0 2 4 6 8 Y-hat 10 p.3.c. Based on the fitted curve and the residual plot, do you feel the model fits well? Fantastic / Pretty Good / Not so good. Q.4. An Analysis of Covariance was conducted to compare the effects of an experimental treatment (use of interactive vocabulary learning) in secondary school German students learning a second language. In the Experimental condition, children made a video fashion show, while in the Control condition, the children completed worksheet assignments. The author fit the following two models, where Y represents post-treatment score, X represents pre-treatment score, and Z1 is 1 if student was in the experimental group, and 0 if in control group. Students were randomly assigned to treatment condition, with nE = 142 and nC = 130. The models fit are: Model 1: E(Y) = X Model 2: E(Y) = X + 1Z1 Model 1: Pre-test score only predictor: Y 8.88 0.70 X SSR1 9145 SSE1 8006 Model 2: Pre-test score and Experimental Condition Dummy Variable (Z1=1 if Experimental, 0 if Control): Y 7.17 0.73 X 2.54Z1 SSR2 9564 SSE2 7587 p.4.a. Test H0: No treatment effect, after controlling for baseline score (1 = 0) HA: 1 ≠ 0 p.4.a.i. Test Statistic: p.4.a.ii. Reject H0 if the test statistic is larger/smaller than ______________________________________ p.4.b. The following table gives the sample means for X and Y by treatment and overall. Compute the adjusted means for the experimental and control groups. Experimental Control Overall X-Mean Y-Mean 11.83 18.37 15.26 18.34 13.47 18.36 Adjusted Mean: Experimental __________________________ Control ___________________________________