Statistical Testing

Statistical significance in all cases is tested by an iterative leave-one-out approach, similar to the analysis of variance (ANOVA) approach used for linear models. For a given set of predictors, the predictive coefficients are calculated to minimize the deviance. The log-likelihood ratio test can then be performed by calculating the difference of deviance for the full model and a reduced model where a single predictor has been left out. This ratio follows a Chi-squared distribution, with degrees of freedom equal to the number of additional parameters in the full model (eg: 1 for speed modulation or a look-up ratemap, 5 for 2D position, etc), giving a p-value. We perform this nested test leaving out each predictor, leading to a complete table of significant predictors.