Table 6.

Goodness of fit metrics for models fit to explain the magnitude of peak activity within each phase (activity_p)

Formula	R²	Root mean squared error	Sigma	AIC
activityp∼(1\|cell)	0.16	0.20	0.21	−527.77
activityp∼phase+s(log2(speed)∗phase)	0.03	0.19	0.19	−1,059.27
activityp∼phase+s(log2(speed)∗phase)+(1\|cell)	0.21	0.17	0.17	−1,317.23

Data come from only the direction closest to the preferred direction for each cell, at each speed. Three GAM models were fit, with potential explanatory variables including log₂ of stimulus speed (speed), phase category (phase; initial transient, transitional, or steady state), and cell identity (cell; as a random effect). Sigma column denotes the residual standard deviation. The best-fitting model, determined by lowest AIC score, is in bold.