dynConfiR (version 0.0.4)

simulateRM: Simulation of confidence ratings and RTs in race confidence models

Description

Simulates the decision responses, reaction times and state of the loosing accumulator together with a discrete confidence judgment in the independent and partially anti-correlated race model (IRM and PCRM) (Hellmann et al., 2023), given specific parameter constellations. See RaceModels for more information about parameters. Also computes the Gamma rank correlation between the confidence ratings and condition (task difficulty), reaction times and accuracy in the simulated output. Basically, this function is a wrapper for rIRM and rPCRM for application in confidence experiments with manipulation of specific parameters. rRM_Kiani simulates a different version of race models, presented in Kiani et al. (2014), but without a confidence measure.

Usage

simulateRM(paramDf, n = 10000, model = "IRM", time_scaled = FALSE,
  gamma = FALSE, agg_simus = FALSE, stimulus = c(1, 2), delta = 0.01,
  maxrt = 15, seed = NULL)
rRM_Kiani(paramDf, n = 10000, time_scaled = FALSE, gamma = FALSE,
  agg_simus = FALSE, stimulus = c(1, 2), delta = 0.01, maxrt = 15,
  seed = NULL)

Value

Depending on gamma and agg_simus.

If gamma is FALSE, returns a data.frame with columns: condition, stimulus, response, correct, rt, conf (the continuous confidence measure) and rating (the discrete confidence rating) or (if agg_simus=TRUE): condition, stimulus,response, correct, rating and p (for the probability of a response and rating, given the condition and stimulus).

If gamma is TRUE, returns a list with elements: simus (the simulated data frame) and gamma, which is again a list with elements condition, rt and correct, each a tibble with two columns (see details for more information).

Arguments

paramDf

a list or data frame with one row. Column names should match the names of RaceModels parameter names (only mu1 and mu2 are not used in this context but replaced by the parameter v). For different stimulus quality/mean drift rates, names should be v1, v2, v3,.... Different s parameters are possible with s1, s2, s3,... with equally many steps as for drift rates. Additionally, the confidence thresholds should be given by names with thetaUpper1, thetaUpper2,..., thetaLower1,... or, for symmetric thresholds only by theta1, theta2,....

n

integer. The number of samples (per condition and stimulus direction) generated. Total number of samples is n*nConditions*length(stimulus).

model

character scalar. One of "IRM" or "PCRM". ("IRMt" and "PCRMt" will also be accepted. In that case, time_scaled is set to TRUE.)

time_scaled

logical. Whether a time_scaled transformation for the confidence measure should be used.

gamma

logical. If TRUE, the gamma correlation between confidence ratings, rt and accuracy is computed.

agg_simus

logical. Simulation is done on a trial basis with RTs outcome. If TRUE, the simulations will be aggregated over RTs to return only the distribution of response and confidence ratings. Default: FALSE.

stimulus

numeric vector. Either 1, 2 or c(1, 2) (default). Together with condition represents the experimental situation. In a binary decision task the presented stimulus belongs to one of two categories. In the default setting trials with both categories presented are simulated but one can choose to simulate only trials with the stimulus coming from one category (each associated with positive drift in one of two accumulators).

delta

numerical. Size of steps for the discretized simulation (see details).

maxrt

numerical. Maximum reaction time to be simulated (see details). Default: 15.

seed

numerical. Seeding for non-random data generation. (Also possible outside of the function.)

Author

Sebastian Hellmann.

Details

The simulation is done by simulating normal variables in discretized steps until one process reaches the boundary. If no boundary is met within the maximum time, response is set to 0. The output of the fitting function fitRTConf with the respective model fits the argument paramDf for simulation. The Gamma coefficients are computed separately for correct/incorrect responses for the correlation of confidence ratings with condition and rt and separately for conditions for the correlation of accuracy and confidence. The resulting data frames in the output thus have two columns. One for the grouping variable and one for the Gamma coefficient.

References

Hellmann, S., Zehetleitner, M., & Rausch, M. (2023). Simultaneous modeling of choice, confidence and response time in visual perception. Psychological Review 2023 Mar 13. doi: 10.1037/rev0000411. Epub ahead of print. PMID: 36913292.

Kiani, R., Corthell, L., & Shadlen, M.N. (2014) Choice certainty is informed by both evidence and decision time. Neuron, 84(6), 1329-1342. doi:10.1016/j.neuron.2014.12.015

Examples

# Examples for "PCRM" model (equivalent applicable for "IRM" model)
# 1. Define some parameter set in a data.frame
paramDf <- data.frame(a=2,b=2, v1=0.5, v2=1, t0=0.1,st0=0,
                      wx=0.6, wint=0.2, wrt=0.2,
                      theta1=4)

# 2. Simulate trials for both stimulus categories and all conditions (2)
simus <- simulateRM(paramDf, n=30,model="PCRM", time_scaled=TRUE)
head(simus)
# equivalent:
simus <- simulateRM(paramDf, model="PCRMt")
# \donttest{
  library(ggplot2)
  simus <- simus[simus$response!=0,]
  simus$rating <- factor(simus$rating, labels=c("unsure", "sure"))
  ggplot(simus, aes(x=rt, group=interaction(correct, rating),
                    color=as.factor(correct), linetype=rating))+
    geom_density(size=1.2)+
    facet_grid(rows=vars(condition), labeller = "label_both")
# }

# automatically aggregate simulation distribution
# to get only accuracy x confidence rating distribution for
# all conditions
agg_simus <- simulateRM(paramDf, n = 20, model="PCRMt", agg_simus = TRUE)
head(agg_simus)
# \donttest{
  agg_simus$rating <- factor(agg_simus$rating, labels=c("unsure", "sure"))
  library(ggplot2)
  ggplot(agg_simus, aes(x=rating, group=correct, fill=as.factor(correct), y=p))+
    geom_bar(stat="identity", position="dodge")+
    facet_grid(cols=vars(condition), labeller = "label_both")
# }

# \donttest{
  # Compute Gamma correlation coefficients between
  # confidence and other behavioral measures
  # output will be a list
  simu_list <- simulateRM(paramDf, model="IRMt", gamma=TRUE)
  simu_list
# }