Package index
Families
One recursion engine, eight models. Each parameter is an ordinary distributional parameter with its own linear predictor, so a learning rate takes a condition effect, a smooth term or a correlated per-subject random effect the way a mean does. Seven turn a value store into a softmax over options; rlddm() turns it into the drift rate of a diffusion, so choices and response times are one likelihood. Every family takes session =, which starts the value store again at each session boundary.
-
bandit2arm_delta() - Rescorla-Wagner delta learning on a two-armed bandit
-
bandit2arm_dual() - Two-armed delta learning with separate rates for gains and losses
-
prl_fictitious() - Counterfactual (fictitious) updating on two options
-
bandit4arm2_kalman_filter() - A Kalman filter over the restless four-armed bandit
-
ts_par7() - The two-step task: a model-based and model-free hybrid
-
igt_pvl_delta() - Prospect-valence learning with a delta rule, for the Iowa gambling task
-
igt_orl() - Outcome-representation learning for the Iowa gambling task
-
rlddm() - Reinforcement learning with a drift-diffusion choice rule
After the fit
These families declare no mean on the response scale, so fitted() and residuals() refuse: for seven of them because the response is a nominal option code, and for rlddm() because the mean of a first-passage time belongs to frmtmb.eam. The trajectory is read here instead.
-
frm_value_trace() - Per-trial value estimates, prediction errors and choice probabilities
-
frm_learn_families() - The families this package supplies, and what each one is
Simulation
Task designs with the payoff schedule of every option fixed in advance, and draws from the generative process.
-
frm_task_design() - Build a trial-level design for a learning task
-
frm_task_simulate() - Draw datasets from a learning family's generative process
-
frmtmb.learnfrmtmb.learn-package - frmtmb.learn: Reinforcement-Learning Families for 'frmtmb' Models