README.md

In psuthaharan/twochoiceRL: Modeling Behavior for Two-Choice Decisions

Modeling two-choice decision behavior in R

See Suthaharan, P., Corlett, P.R., & Ang, Y.S. (2021). Computational modeling of behavioral tasks: An illustration on a classic reinforcement learning paradigm for more detail.

Questions? Contact Praveen Suthaharan ([e-mail]).

—————

The goal of twochoiceRL is to guide users on how to simulate two-choice decision behavior and then how to recover the behavioral parameters that were used to simulate the behavior.

Installation

You can install the development version of twochoiceRL from GitHub with:

# install.packages("devtools")
devtools::install_github("psuthaharan/twochoiceRL")

Simulation

Simulate behavior:

library(twochoiceRL)
# generate 100 individuals
# dataset1 contains behavioral data keeping only those selected-choice trials
dataset1 <- simulate_twochoiceRL(n_subj = 100, n_tr = 200, trials_unique = TRUE)

# dataset2 contains behavioral data keeping all trials (selected-choice and the non-selected-choice)
dataset2 <- simulate_twochoiceRL(n_subj = 100, n_tr = 200, trials_unique = FALSE)

Visualization

Plot behavior:

# for plotting purpose - let's use dataset2
# You randomly select a participant to observe his/her behavior

# View first 10 rows of individual 100's data
head(dataset2$twochoiceRL[[100]],10)
#>    Trial        Value        Pr Option Action      Reward
#> 1      1  0.011472946 0.5000000      1      1  0.07437611
#> 2      1  0.000000000 0.5000000      2      1  0.07437611
#> 3      2  0.150651408 0.5024829      1      1  0.91373052
#> 4      2  0.000000000 0.4975171      2      1  0.91373052
#> 5      3  0.150651408 0.5325572      1      2 -1.39233659
#> 6      3 -0.214775990 0.4674428      2      2 -1.39233659
#> 7      4  0.028651131 0.5784314      1      1 -0.64024449
#> 8      4 -0.214775990 0.4215686      2      1 -0.64024449
#> 9      5 -0.001389443 0.5524874      1      1 -0.16609406
#> 10     5 -0.214775990 0.4475126      2      1 -0.16609406


# Visualize behavior 
p1 <- plot_twochoiceRL(data = dataset2, subj = 100, colors = c("#009999","#0000FF"))

# View expected value AND choice probability
p1[[3]]

# Visualize choice probability - animated 
plot_twochoiceRL(data = dataset2, subj = 100, colors = c("#009999","#0000FF"), plot_type = "animate")

Estimation

Estimate behavioral parameters from simulated data:

# Run MLE (elapsed time: ~ 5 min)
est_mle <- estimate_twochoiceRL(data = dataset1,
                                method = "mle",
                                plot = TRUE)
est_mle[[4]] # return MLE-estimation plot

# Run MAP (elapsed time: ~ 5 min)
est_map <- estimate_twochoiceRL(data = dataset1,
                                method = "map",
                                prior_mean = c(0,0),
                                prior_sd = c(5,5),
                                plot = TRUE)
est_map[[4]] # return MAP-estimation plot

# Run EML (elapsed time: ~ 1.4 hours)
est_eml <- estimate_twochoiceRL(data = dataset1,
                                method = "eml",
                                prior_mean = c(0,0),
                                prior_sd = c(5,5),
                                plot = TRUE)
est_eml[[4]] # return EML-estimation plot

Visualize the posterior hyperparameters per EML-iteration:

# Laplace-approximated posterior hyperparameters
# for x1 AND x2
est_eml[[7]]

# x1, animated
est_eml[[8]]

psuthaharan/twochoiceRL documentation built on Dec. 22, 2021, 9:56 a.m.