rtdists (version 0.2-6)

LBA-race: LBA race functions: Likelihood for first accumulator to win.

Description

n1PDF and n1CDF take RTs, the distribution functions of the LBA, and corresponding parameter values and put them throughout the race equations and return the likelihood for the first accumulator winning (hence n1) in a set of accumulators.

Usage

n1PDF(t, A, b, t0, ..., st0 = 0, distribution = c("norm", "gamma",
  "frechet", "lnorm"), args.dist = list(), silent = FALSE)

n1CDF(t, A, b, t0, ..., st0 = 0, distribution = c("norm", "gamma",
  "frechet", "lnorm"), args.dist = list(), silent = FALSE)

Arguments

t

a vector of RTs.

A,b

LBA parameters, see LBA. Can either be a single numeric vector (which will be recycled to reach length(t) for trialwise parameters) or a list of such vectors in which each lis

t0

one scalar t0 parameter (see LBA). Multiple or trialwise t0 parameters are currently not implemented.

...

two named drift rate parameters depending on distribution (e.g., mean_v and sd_v for distribution=="norm"). The parameters can either be given as a numeric vector or a list. If a numeric vector

st0

one scalar parameter specifying the variability of t0 (which varies uniformly from t0 to t0 + st0).

distribution

character specifying the distribution of the drift rate. Possible values are c("norm", "gamma", "frechet", "lnorm"), default is "norm".

args.dist

list of optional further arguments to the distribution functions (i.e., posdrift or robust for distribution=="norm").

silent

logical. Should the number of accumulators used be suppressed? Default is FALSE which prints the number of accumulators.

Details

For a set of $N$ independent accumulators $i = 1...N$, the race likelihood for a given accumulator $i$ is given by $$L(\mbox{unit }i \mbox{ wins}) = f_i(t) \times \prod_{j<>i} [ S_j(t) ]$$ where $f(t)$ is the PDF (dlba_...) and $S_j(t) = 1 - F_j(t)$ is the survivor function, that is the complement of the CDF $F(t)$ (plba_...) at time $t$. In other words, this is just the PDF/CDF for the winning accumulator at time $t$ times the probability that no other accumulators have finished at time $t$.

Examples

## check random generated values against race functions:

## 1. Without st0:
r_lba <- rlba_norm(1e4, A=0.5, b=1, t0 = 0.5, mean_v=c(1.2, 1), sd_v=0.2)
x <- seq(0.5, 4, length.out = 100) # for plotting
# PDF
y <- n1PDF(x, A=0.5, b=1, t0 = 0.5, mean_v=c(1.2, 1.0), sd_v=0.2) # PDF
hist(r_lba$rt[r_lba$response==1],probability = TRUE, breaks = "FD")
lines(x=x,y=y/mean(r_lba$response == 1))
# CDF
plot(ecdf(r_lba$rt[r_lba$response==1]))
y <- n1CDF(x, A=0.5, b=1, t0 = 0.5, st0 = 0, mean_v=c(1.2, 1.0), sd_v=0.2)
lines(x=x,y=y/mean(r_lba$response == 1), col = "red", lwd = 4.5, lty = 2)
# KS test
normalised_n1CDF = function(t,...) n1CDF(t,...)/n1CDF(t=Inf,...) 
ks.test(r_lba$rt[r_lba$response==1], normalised_n1CDF, A=0.5, b=1, t0 = 0.5, 
        mean_v=c(1.2, 1.0), sd_v=0.2)

## 2. With st0 = 0.2:
r_lba <- rlba_norm(1e4, A=0.5, b=1, t0 = 0.5, st0 = 0.2, mean_v=c(1.2, 1), sd_v=0.2)
x <- seq(0.5, 4, length.out = 100) # for plotting
# PDF
y <- n1PDF(x, A=0.5, b=1, t0 = 0.5, st0 = 0.2, mean_v=c(1.2, 1.0), sd_v=0.2) # PDF
hist(r_lba$rt[r_lba$response==1],probability = TRUE, breaks = "FD")
lines(x=x,y=y/mean(r_lba$response == 1))
# CDF
plot(ecdf(r_lba$rt[r_lba$response==1]))
y <- n1CDF(x, A=0.5, b=1, t0 = 0.5, st0 = 0.2, mean_v=c(1.2, 1.0), sd_v=0.2)
lines(x=x,y=y/mean(r_lba$response == 1), col = "red", lwd = 4.5, lty = 2)
# KS test
normalised_n1CDF = function(t,...) n1CDF(t,...)/n1CDF(t=Inf,...) 
ks.test(r_lba$rt[r_lba$response==1], normalised_n1CDF, A=0.5, b=1, t0 = 0.5, 
        st0 = 0.2, mean_v=c(1.2, 1.0), sd_v=0.2)

## Other examples:

xx <- rlba_norm(10, A=0.5, b=1, t0 = 0.5, mean_v=1.2, sd_v=0.2)

# default uses normal distribution for drift rate:
n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, mean_v=c(1.2, 1.0), sd_v=0.2)

# other distributions:
n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, shape_v=c(1.2, 1), scale_v=c(0.2,0.3), distribution = "gamma")
n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, shape_v=c(1.2, 1), scale_v=c(0.2,0.3), distribution = "frechet")
n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, meanlog_v = c(0.5, 0.8), sdlog_v = 0.5, distribution = "lnorm")

# add st0:
n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, mean_v=c(1.2, 1.0), sd_v=0.2, st0 = 0.4)


# use different A parameters for each RT:
n1PDF(xx$rt, A=runif(10, 0.4, 0.6), 
      b=1, t0 = 0.5, mean_v=c(1.2, 1.0), sd_v=0.2)

# use different A parameters for each RT and each accumulator:
n1PDF(xx$rt, A=list(runif(10, 0.4, 0.6), runif(10, 0.2, 0.4)), 
      b=1, t0 = 0.5, mean_v=c(1.2, 1.0), sd_v=0.2)


### vectorize drift rates:

# vector versus list:
v1 <- n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, mean_v=c(1.2, 1.0), sd_v=0.2)
v2 <- n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, mean_v=list(1.2, 1.0), sd_v=0.2)
identical(v1, v2)  # TRUE

# drift rate per trial:
n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, mean_v=list(rnorm(10, 1.2), rnorm(10, 1)), sd_v=0.2)

# combine list with vector:
n1PDF(xx$rt, A=0.5, b=1, t0 = 0.5, mean_v=list(rnorm(10, 1.2), rnorm(10, 1)), sd_v=c(0.2, 0.1))