Trend step size #

The increment (in units of years) by which the hat-type means are evaluated.

Syntax #

trend step size: 1 # (0, 1]

Usage #

The default trend step size is 0.2 years, which means we get five predictions in every calendar year. For instance, between years 0 and 1, this equates to predictions at

0.0 0.2 0.4 0.6 0.8 1.0

This yields hat-type means that appear very smooth, especially when the trend is curvilinear. Increasing the step size saves computation time (and the size of in-memory JAGS objects), but at the cost of a smooth trend line. Consider the following:

The teal line corresponds to a trend step size of 1, while the salmon-colored line corresponds to the default trend step size. The larger trend step size gives the false impression that the change in the mean is stepwise linear, when in fact it is not. If using a larger trend step size to fit “lighter” models more quickly, this is worth bearing in mind.