Skip to contents

Per-participant information-value (infoVal) summary

Source: R/compute_infoval_summary.R
compute_infoval_summary.Rd

Computes the information value (infoVal; Brinkman et al., 2019) for every participant and returns a pass / warn / fail summary plus a per-participant table flagging those below the threshold.

Usage

compute_infoval_summary(
  responses,
  method = c("2ifc", "briefrc"),
  rdata,
  baseimage = "base",
  col_participant = "participant_id",
  col_stimulus = "stimulus",
  col_response = "response",
  iter = 10000L,
  threshold = 1.96,
  ...
)

Arguments

responses

A data frame of trial-level responses.

method

"2ifc" (supported) or "briefrc" (returns "skip").

rdata

Path to the rcicr .RData file that produced the stimuli. Required for 2IFC.

baseimage

Name of the base image used at generation time (the key in base_face_files in the rdata). Default "base".

col_participant, col_stimulus, col_response

Column names.

iter

Number of iterations for rcicr's reference-distribution simulation. Default 10000 (the rcicr-recommended value).

threshold

Numeric. Participants with infoVal below this are flagged as likely noise. Default 1.96.

...

Unused.

Value

An rcdiag_result() object. For 2IFC, data$per_participant has one row per participant with participant_id, infoval, and meaningful (logical: infoval >= threshold). For Brief-RC, a "skip" result with an explanatory message.

Details

infoVal is a z-like score describing how far a participant's classification image (CI) is from a null-response reference distribution. Values at or below 1.96 are effectively indistinguishable from noise; higher values indicate meaningful signal.

The 2IFC path delegates to rcicr::batchGenerateCI2IFC() and rcicr::computeInfoVal2IFC() from the canonical rcicr package (Dotsch, 2016; v1.0.1 on GitHub as of 2023). Canonical rcicr does not expose Brief-RC-specific CI or infoVal functions, so the Brief-RC path returns a "skip" result. A correct Brief-RC infoVal requires a reference distribution matched to each participant's trial count (not the pool size stored in the rdata), and implementing that correctly is deferred to the companion rcicrely package.

Side effect (2IFC). rcicr caches a reference distribution inside the supplied rdata file on the first call. Subsequent calls reuse it. Copy your rdata beforehand if you want the original untouched.

References

Brinkman, L., Goffin, S., van de Schoot, R., van Haren, N. E., Dotsch, R., & Aarts, H. (2019). Quantifying the informational value of classification images. Behavior Research Methods, 51(5), 2059–2073.

Dotsch, R. (2016). rcicr: Reverse-correlation image-classification toolbox [R package]. https://github.com/rdotsch/rcicr

Examples

if (FALSE) { # \dontrun{
compute_infoval_summary(
  responses, method = "2ifc",
  rdata = "stimuli.RData",
  iter = 10000
)
} # }