In Figure 61 we overlay on the scatterplot of the raw data a loess smooth and the logistic regression fits given by
In comparing the linear model with the loess fit we see that for values of Age > 50 the loess fit and the logistic regression line don't match well. In addition, for small values of Age (near zero) and for values of Age ≈ 15 the logistic fit does not match the data well. This mismatch of the logistic regression fit and the data may indicate that adolescents may have been more likely to survive. When we compare the quadratic model with the data we see that it fits better for Age ≈ 15 but performs much poorly for Age35 where the quadratic model predicts a zero probability of survival.
