STA 242/ENV 255: January 28, 1998

Polynomial and Curvilinear Regression

Assignment: Due Tuesday -- February 3 -- by 5pm

Suggested Reading: LAB GOALS: Fit curves to data through either (1) transformed X and/or Y variables (Problems 2:11-13) or by (2) Polynomials (Problem 5:5-7)

SAS datasets

Create SAS datasets forProblems 9-13, page 61 and Problems 5-7, page 176

Refer back to last week's lab for help creating data sets. Make sure that the variable names are 8 characters or less!

Suggestion: Try to create the data sets before lab.

Transformations

SAS/INSIGHT has several common transformations available through the EDIT menu. (For a complete discussion, see chapter 20 pages 241-256 of the SAS/INSIGHT manual). For example, you are asked to transform the data using the log base 10 of YIELD. Pull down the Edit menu: 
 
 
Edit
Variables ->
log()
In the popup window, select log10(Y+a) for the transformation and click on Yield and then Y. The default name is L_Yield; modify the Name: box if you want to use a different name. Click on OK and the column L_YIELD is added to your spreadsheet.

Polynomial Regression

SAS INSIGHT Chapter 13. pages 155-169.

Select Analyze:Fit from the DATA window. Put Location in the Label region , D in the Y region, and Latitude in the X region. This fits the usual straight line regression. What do the residual plots, t-tests and R2 suggest about the fit of the model?

To add polynomial terms, go to the box in the ouput labeled "Parametric Regression". In the slider under Degree (Polynomial) click on the right arrow until the degree is 2. (Quadratic regression). You may want to experiment by adding more terms.... What do the residual plots suggest about the fit of the quadratic model - are you happy accepting this model or should we be concerned?

For Printing output, use the File:Print... menu File:Print... menu