ch 8 logistic regression

8.1 logistic regression and a single predictor
    (binomial logit model I think)

8.1.1 logit function and log-odds

8.1.2 likelihood

8.1.3 deviance (G^2, n - #betas, null & residual deviance, deviance
residuals)

8.1.4 comparing with deviance

       - changes in deviance are LR tests

       - pearson GOF test

8.1.5 R^2){dev}...  1 - (HA misfit)/(H0 misfit); there are other
definitions, but again the idea is to show % reduction in misfit from
the HA model

8.1.6 residuals

       - response residuals

       - pearson residuals (stdized residuals)

       - deviance residuals

       - stdized dev residuals (divide by sqrt(1-hii)

8.2 binary logistic regression  [and why different from binomial logit model]

8.2.1 binary logistic regression deviance

8.2.2 residuals for binary data

8.2.3 X transformations,

    X dist  log odds(x)
    normal  quadratic [linear spec case]
    poiss   linear
    dummy   linear
    skewed  incl x and log(x) in the model

8.2.4 marg mod plots