2.4 The change of skull size from 1953 to 2004
We first analysed the changes in the overall size of S. araneususing the full dataset from 1953-2004, which was only available for
juveniles in July and adults. We tested the significance of the slopes
of the linear regressions of all three dimensions over time: skull
height, width, and length. We added a quadratic term to the linear model
to test for linearity of the relationships and determine whether it
improved the model according to an F test with a type I sum of squares.
However, inspection of the residuals from the linear regression
suggested a more subtle nonlinearity of the skull size over time. Thus,
we additionally used locally weighed regression smoothing (LOESS;
Cleveland 1979). We selected the smoothing parameter by minimizing
Akaike’s information criterion, which strikes a balance between the
residual sum of squares and the complexity of the fit.