Extended Data Fig. 6: Varying dimensionality of subspaces.
From: Separating cognitive and motor processes in the behaving mouse
Analyses were repeated while varying the dimensionality of movement-null and movement-potent subspaces. Each subspace was constrained to be 4 (left), 6 (middle left), 8 (middle right), or 13 dimensions (right). a. Upper bound estimate of dimensionality for trial-averaged (PSTH) data and single-trial data. Bar heights indicate mean across sessions, points indicate sessions, and error bars indicate standard deviation across sessions (n = 25 sessions). b. Cumulative variance explained of the neural activity by the activity in movement-null and movement-potent subspaces. Bold lines and points indicate mean across sessions. Thin lines represent single sessions c. Normalized variance explained of neural activity during the delay or response epoch by the activity in movement-null and movement-potent subspaces. Points indicate sessions, bar height indicates mean across sessions, and error bars indicate standard deviation across sessions (n = 25 sessions). d–f. Subspace (d), CDchoice (e), and CDramp (f) alignment distributions when varying dimensionalities of each subspace.
