This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: PUBH 7430 Lecture 9 J. Wolfson Division of Biostatistics University of Minnesota School of Public Health October 4, 2011 The Generalized Linear Model The Generalized Linear Model generalizes the standard linear model by: 1 Letting a function of the mean μ be a linear combination of the covariates: g ( μ ) = X β 2 Letting the variance be a function of the mean: Var ( Y  X ) = V ( μ ) GLMs: Coefficient interpretation Coefficient interpretation must take into account the link function g : “A oneunit increase in X is associated with a βunit increase/decrease in g ( μ )” Can work backwards to get familiar cases: • Poisson regression: “A oneunit increase in X is associated with an exp( β ) percent increase/decrease in the mean of Y ” • Logistic regression: “A oneunit increase in X is associated with an exp( β ) 1+exp( β ) percent increase/decrease in the odds of Y ” Aside: Log transformation • When data are positive and display evidence of nonnormality, log transformation is often recommended before proceeding with inference. • If using a linear model, we are assuming: E [log( Y i )  x i ] = x i β Var [log( Y i )  x i ] = σ 2 • An alternative is to fit a GLM with a log link and our choice of variance function: log( E [ Y i  x i ]) = x i β Var [ Y i  x i ] = V ( μ i ) Aside: Log transformation...
View
Full
Document
 Fall '04
 Prof.Eberly

Click to edit the document details