Exercise 28.4  of PSLS 3e:
-------------------------

A study on nestling mass (grams) of great titmice parents and its
association with nest humidity index and exposure to fleas during
egg laying (a categorical and binary predictor, coded as 0/1).
The model fitted was a multiple linear regression model with two predictors, 

  Y_i = beta0 + beta1*Humidity_i + beta2*Exposed_i + eps_i, i=1,...,37.

(a) The estimated regression was: 

  Nestling mass = 18.085 - 5.411*Humidity + 0.848*Exposed.


(b) The coefficient for Exposed was positive (0.848) and weakly significant
(P=0.024). Hence we have evidence to say that the nestling mass
was higher in nests exposed to fleas during egg laying.


(c) The regression standard error (the somewhat unusual and perhaps
confusing term used in the chapter for the residual standard deviation)
equals 1.016. It is the standard deviation in the normal distribution of
the errors, or the unexplained variation in the model.


(d) R2 is given as 47.7%. This means that the regression equation
explains roughly half of the variation in the data (among the nestling
masses). It's a quite decent value for a model with only two parameters,
and indicates that these two parameters do a good job of explaining
the variation; the listing shows that it's in particular the Humidity
that has a strong effect. The moderate R2 value shows prediction errors 
will be moderate but not very small (or very large). The residual 
standard deviation gives a quantitative value for the unexplained variation.