For the horseshoe crab data introduced in Section 1.5.1 and modeled in Section 7.5, suppose you use graphics to investigate how a female crab’s number of male satellites depends on the width (in centimeters) of the carapace shell of the crab. If you plot the response counts of satellites against width, the substantial variability makes it difficult to discern a clear trend. To get a clearer picture, fit a generalized additive model, assuming a Poisson distribution and using the log link. What does this suggest about potentially good predictors and link functions for a GLM?