Non-linear regression trees
In addition to linear relations, we also investigated non-linear
relationships between measures of inhibitory performance and
questionnaires with regression trees. To test the influence of BIS-11
and OCI-R scores, each model was tested via nested cross-validation and
on a test set of participants from a different sample. Regression trees
from nested cross-validation were also tested in permutation models with
shuffled dependent variables to assess their superiority over a random
model. Model accuracy is reported via root mean squared error. As is
evident from Table 5, the models differ only marginally from one another
in terms of accuracy. While the regression tree seems to explain IES
better than the linear model on the test set, this does not hold true
for our sample data in nested cross-validation. Therefore, non-linear
associations between impulsivity and compulsivity with inhibitory
performance were not observed.