In SpTB/wztools: Convenient functions for analysis and simulation

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(wztools)
library(magrittr)

For this vignette, we will use 2 attached synthetic datasets. First, for a single subject:

single_subj_sim

This is a simulated dataset of accuracies and reaction times in two difficulty conditions (see ?single_subj_sim). We will assume that conditions are blocked. We can use the make_stan_list function to transform the output to stan-appropriate format:

var_names = c('choice', 'rt') #specifies which variables we want to summarise in multi-dimentional arrays
grouping_vars = c('cond', 'trial') #specifies by which dimentions will the output be grouped
grouping_vars_stan_names=c('numConds','numTrials')
stan_list = make_stan_list(dat = single_subj_sim,
                           var_names = var_names,
                           grouping_vars = grouping_vars,
                           grouping_vars_stan_names = grouping_vars_stan_names,
                           hierarchical = F # non-hierarchical setting is for a single subject
                           )
stan_list

The stan_list elements rt and choice are arrays, with condition and trial being the first and second dimention, respecively.

When using evidence accumulation to model choices, we might need to additionally add the minimal rt (rt bound), which we can do using the ssm_prepro argument:

stan_list = make_stan_list(dat = single_subj_sim,
                           var_names = var_names,
                           grouping_vars = grouping_vars,
                           grouping_vars_stan_names = grouping_vars_stan_names,
                           hierarchical = F, # non-hierarchical setting is for a single subject
                           ssm_prepro=T)
stan_list

We can do this also for multisubject data:

#pre-generated 10-subject dataset
sim

We just need to add subject identifier (subjID) as another dimension to our arrays. We can do this by adding subjID as the first grouping variable, and changing the hierarchical argument to TRUE:

grouping_vars = c('subjID', 'cond', 'trial') 
grouping_vars_stan_names=c('numSubj', 'numConds','numTrials')

stan_list = make_stan_list(dat = sim,
                           var_names = var_names,
                           grouping_vars = grouping_vars,
                           grouping_vars_stan_names = grouping_vars_stan_names,
                           hierarchical = T, # non-hierarchical setting is for a single subject
                           ssm_prepro=T)
dim(stan_list$choice)

Now choice and rt are 3-dimentional arrays, with subjID being the first dimension. Such multi-dimensional arrays are a convenient way of inputting data into Stan.

SpTB/wztools documentation built on July 2, 2023, 11:50 a.m.