CAS MA 416 Lecture 4: MA416 Class 4
Document Summary
Ma416 class 4: the multiple linear regression model. Independent variables: measurement variable (or dummy/indicator variables more on this in 2 weeks) Note: k independent variables, (k+1) parameters in the model. The multiple regression model is a conditional model, in that i describes the effect of xi , if all other independent variables in the model remain constant. Prediction, more predictors lead to more accurate prediction. Identify set of important predictors from a larger set of potential predictors. Separate out the effects of different predictors, control for confounding. Least squares criteria choose parameter estimates that minimize the sum of squared errors, The least squares estimates also maximize the correlation between y and y (over all possible linear models). To fit the model, need more data than parameters. A common guideline for a minimal practical sample size is to have at least 10 observations for each variable in the model. Linearity also means linear in the regression coefficients.