problem-3.25

problem-3.25  The wtloss data can be plotted as follows:
> attach(wtloss)
> cor(Weight, Days)             # close to -1
[1] -0.9853
> plot(Days, Weight)            # kinda linear
> res = lm(Weight ~ Days)
> abline(res)                   # add regression line
> plot(Days, residuals(res))    # A trend
> detach(wtloss)
    
We see from the scatterplot that a linear model may not explain the data well, despite the correlation coefficient that is close to -1. The residual plot amplifies this observation. This is to be expected as it is initially easy to lose a large amount of weight, but this gets progressively harder as there is less excess weight to shed.