Overview
Upload a single MS-Word document with your answers.
Use the HBAT_200.sav file. Recode ID into a different variable named Split130 with ID numbers 1-130 having a value of 0 and 131-200 having a value of 1.
1) Multiple Discriminant Analysis
Stage 1: Objectives
Research Situation: Interested in any differences in perceptions between customers located/served by their USA-based salesforce versus those outside the United States.
Stage 2: Research Design
Variables Used
Dependent Variable – Region (X4)
Independent Variables – 13 perception variables (X6 to X18)
Sample of 200 observations
1a. What are the group sizes for the estimation sample and the validation sample?
Stage 3: Assumptions
1b. What is the significance level for the Box’s M for the discriminant model using the stepwise method and Mahalanobis distance? Are the covariance matrices the same?
Stage 4: Estimation of the Discriminant Model and Assessing Overall Fit
1c. What independent variables have significant difference in their group means using a 5% level of significance?
1d. What are the standardized canonical discriminant function coefficients?
1e. In terms of overall fit, what percentage of the variance in X4 is explained by this model?
1f. What is the prediction accuracy of the estimation sample? What is the prediction accuracy of the holdout sample? What is the proportional chance value? What is the maximal chance value?
1g. What is the most important independent variable based on the discriminant loadings in the structure matrix?
2) Logistic Regression Analysis
Stage 1: Objectives
Research Situation: Interested in any differences in perceptions between customers located/served by their USA-based salesforce versus those outside the United States.
Stage 2: Research Design
Variables Used
Dependent Variable: Region (X4), Independent Variables: 13 perception variables (X6 to X18)
Sample Size
Sample of 200 observations
Stage 3: Assumptions of Logistic Regression
Minimal assumptions (independence of observations) met.
Stage 4: Estimation of the Logistic Regression Model and Assessing Overall Fit
Use the stepwise Forward-Wald Method with Split 130 as the Selection Variable, Value = 0 and 100 maximum iterations.
Overall Model Fit
2a. What is the Cox & Snell R-Square and the Nagelerke R-Square for the final model?
Overall Predictive Accuracy:
2b. What is the overall percentage correct for the estimation sample using a 0.5 cut value?
2c. What is the significance level for the Hosmer and Lemeshow Statistical Test? Does the model fit well?
2d. What is the area under the ROC curve for Predicted probability (PRE_1) for the estimation sample?
2e. Paste a Scatter Plot of the Normalized residual on the Y-axis and ID on the X-axis for the estimation sample. What ID numbers have values above 2?
Stage 5: Interpreting the Logistic Coefficients
2f. What variables are statistically significant in the final model?
2g. Which (if any) of the significant variables has a negative relationship to the outcome (X4 = 1 = outside North America)?
Stage 6: Validation of the Results
2h. What is the overall percentage correct for the holdout sample using a 0.5 cut value? How does that compare to the proportional chance value?