ISyE 3030 Basic Statistical Methods Other topics about regression (chp 11, 12) Instructor: Professor Jing Li H. Milton Stewart School of Industrial and Systems Engineering Georgia Tech 1 Correlation (chp 11.8) 2 Correlation coefficient 3 Hypothesis Test • Testing whether or not there is • Test statistic 4 Example 11.8 5 Example 11.8 6 Example 11.8 7 Regression on transformed variables (chp 11.9) 8 Regression on transformed variables • Deal with non-linearity: Sometimes visual inspections, or prior knowledge, tells us that there are some non-linear factors in regression model • Examples: 9 Example 11.9: Wind-mill power 10 Try fitting a linear model? • Result of fitting linear regression model Residual • Residual plot indicates the linear relationship does not capture all the information in the wind-speed variable. 11 A second try • As wind speed increases, output (y) approach to an upper limit (consist with physics of windmill operation) Raw data Transformed data 12 Residual: Linear model Residual: Transformed data model 13 14 Logistic regression (chp 11.10) 15 Logistic regression ππππ = π½π½0 + π½π½1 π₯π₯ππ + ππππ πΈπΈ ππππ = 1 1 + ππππππ − π½π½0 + π½π½1 π₯π₯ππ Linear regression does not work because the assumptions are not met. Logistic regression To estimate the logistic regression coefficients, we can use maximum likelihood estimation (details skipped). 16 Example 17 Example 18 Example Fitted logistic regression πΈπΈ ππππ = 1 1 + ππππππ − 10.875 − 0.17132π₯π₯ππ 19 Polynomial regression (chp 12.6.1) 20 Polynomial regression 21 Example 12.12 22 Example 12.12 23 Example 12.12 24 Categorical regressors (chp 12.6.2) 25 Categorical regressors 26 Example 12.13 27 Example 12.13 28 Example 12.13 29