工程建设统计学Handout3.pdfVIP

Engineering Statistics （工程统计学） Handout #3: Linear Regression Analysis A non-probabilistic approach Jiandong Wang jiandong@pku.edu.cn College of Engineering, Peking University Outline Outline • Motivation • Overview of modeling • Curve fitting (Least-squares method 最小二乘方法) • Vector space – Geometric interpretation of the least-squares method • Generalization of the least-squares method Key points ： • Understand the principles and geometric interpretations of the LSM, the PCA-projection method, the locally-weighted LSM, and use these methods to solve the curve fitting problem Assignment #2: see the last slide, Deadline: Oct 15, 2013 References • Chapter 2, Chapter 8, Section 3.4 and Section 3.6 in Hardle and Simar (2007) • Section 11.1, Section 11.2, Section 12.1 in Montgomery and Runger (2011) 2 Motivation Motivation • Given a set of data samples {xi , yi }n , how to fit a line or curve to the data i=1 samples, so that a good estimate of y can be achieved for another value x ? 0 0 Example 1: The grades of 10 students on a midterm exam x and on the final exam y are as follows: x 77 50 71 72 81 85 94 96 99 67 y 82 66 78 64 67 ? 85 99 99 68 Could you provide an estimate of the missing value y for x = 85? Example 2. Given a set of data samples {x , y }n shown here, what is a good i i i=1 estimate of y0 for x0 = 3.5? Lec3_LinearRegresson.m Lec3_NonLinearRegresson.m 3 Motivation Motivation • Modeling is one of the most important techniques in Sta