Lecture14-PennStateDepartmentofStatistics.docVIP

Chapter 6 First- Order Model w/2 Covariates (i.e. Predictors) Model: Yi = Bo + B1Xi1 + B2Xi2 + ei where i = 1,….,n E(Yi) = b0 + b1Xi1 + b2Xi2 and the assumptions are same as SLR: e’s normal, independent and constant variance. B1 = change in mean response E(Y) per unit change in X1 where X2 is held constant. (For models with more than 2 predictors then all other X’s are held constant). B2 = change in mean response E(Y) per unit change in X2 where X1 is held constant. (For models with more than 2 predictors then all other X’s are held constant). e.g. E(Y|X1+1,X2) - E(Y|X1,X2) = b0 + b1(Xi1+1) + b2Xi2 – [b0 + b1Xi1 + b2Xi2] = b0 + b1Xi1 + b1 + b2Xi2 - b0 - b1Xi1 - b2Xi2 = b1 Also X1 and X2 have additive effects on Y (i.e. no interactions) and the plot is a plane. First- Order Model more than 2 Covariates (i.e. Predictors) Model: Yi = Bo + B1Xi1 + B2Xi2 +,….,+Bp-1Xp-1 + ei where i = 1,….,n and p = number of estimates in model. (For example w/ 2 predictors p = 3.) or General Linear Regression Model w/ p-1 covariates and is called a first-order model when X1,…,Xp-1 are different p-1 covariates all raised to a power of one. Qualitative or Dummy Variables Example: Y – shipping cost of express delivery service and there are three Cargo types: Fragile, Semi-fragile, and Durable: and durable is called the “base level”. The values mean nothing, but can change the direction of the relationship. Model: E(Yi) = b0 + b1Xi1 + b2Xi2 Fragile: E(Yi) = b0 + b1(1) + b2(0) = b0 + b1 = uF Semi: E(Yi) = b0 + b1(0) + b2(1) = b0 + b2 = uS Durable: E(Yi) = b0 + b1(0) + b2(0) = b0 = uD Therefore:B1 = uF – uD and B2 = uS - uD This is why Durable is called “base level” since the other cargo types are being compared to durable. This base level is arbitrary as either Fragile or Semi-fragile could have been classified as the base level. Note: in MTB if you would created a dummy variable for Durable, say X3, and entered all three dummy variables into the mode