STAT3012 Lecture Notes - Lecture 12: Numerical Stability, Simple Linear Regression, Correlation And Dependence

• School: University of Sydney
• Department: Statistics
• Course: STAT3012
• Professor: Michael

Document Summary

A polynomial linear regression model of degree k n is. + kxk i + i, (i = 1, . N are random errors satisfying the usual assumptions (a1) (a4). Typically statisticians aim for low order polynomial models having degree k larger than 5 or 6 is unusual. If you use a polynomial model of degree k, then you should include all terms of lower order (unless the circumstances are exceptional). A cubic polynomial should include linear and quadratic terms: E[yi|xi] = 0 + 1xi + 2x2 i + 3x3 i. We should not omit (e. g. ) the quadratic term: E[yi|xi] = 0 + 1xi + 3x3 i. Fev (forced expiratory volume) is a mea- sure of lung function. The data include determinations of fev on 318 female children who were seen in a childhood respiratory disease study in mas- sachusetts in the u. s. Of interest to model fev by age.