2/3/2009

Inferences in Regression and Correlation Analysis (continued)

• Correlation:
• Pearson
• Spearman
• Nonparametric correlation
• Procedure

1. rank within Y1 and Y2

2. Pearson correlation on ranks

• Inference
• Using SAS

Checking Model Assumptions

• Possible departures from the Model:
• Non-linear relationship
• Non-constant variance
• Outliers
• Non-independent error terms
• Non-normal error terms
• Variables that should be included in the model
  
  
• Residuals & Studentized Residuals
  
  
• Example: airfreight.sas

• Tests for Normality
• Correlation test
• Based on expected value of ordered residuals
• Anderson-Darling test (n ≤ 25)
• Shapiro-Wilk test (n ≤ 2000)
• Kolmogorov-Smirnov test (n > 2000)
  
  
• Tests for Constant Variance
• Brown-Forsythe Test (robust)
• Example: Using SAS
• Breusch-Pagan Test (large-sample, normal errors)
• Example: Using SAS

Please read the rest of Chapter 3 for Thursday.