Cross-validate infoVal against RT quality — check_rt_infoval

Correlates per-participant infoVal with per-participant median response time. The check looks for two patterns that are individually plausible but jointly suspicious:

Usage

check_rt_infoval_consistency(
  responses,
  method = c("2ifc", "briefrc"),
  rdata = NULL,
  stimuli = NULL,
  noise_matrix = NULL,
  base_image = "base",
  col_participant = "participant_id",
  col_stimulus = "stimulus",
  col_response = "response",
  col_rt = "rt",
  iter = 1000L,
  seed = NULL,
  ...
)

Arguments

responses: Data frame with one row per trial. Required columns: participant_id, stimulus, response (values in {-1, +1}), and rt (response time, milliseconds). Load yours from CSV via read_responses() or utils::read.csv(); column names are configurable via the col_* arguments.
method: "2ifc" or "briefrc".
rdata: Path to the rcicr .RData file (2IFC). Either rdata or stimuli must be supplied for the 2IFC path.
stimuli: In-memory stimuli list (the $stimuli element of an rcisignal_sim object). Use in place of rdata when the file path no longer resolves (e.g. after saveRDS()/ readRDS() across R sessions).
noise_matrix: Noise matrix for the Brief-RC path. Either a numeric matrix of n_pixels x pool_size, or a path to a text / .rds file (see read_noise_matrix()).
base_image: Name of the base image in rdata$base_face_files (2IFC only). Default "base".
col_participant, col_stimulus, col_response, col_rt: Column names. col_rt is required.
iter: Reference-distribution iterations. Default 1000L.
seed: Optional integer; RNG state restored on exit.
...: Unused.

Value

An rcisignal_diag_result() object. data$per_participant has participant_id, infoval, median_rt. data$correlation is the Pearson correlation between the two.

Details

High infoVal with fast median RT. A participant producing a seemingly informative CI while responding faster than others is more likely to have produced a spurious signal than genuine meaningful information – the signal may be an artefact of a button-mashing strategy that happens to correlate with the noise.
A negative correlation between infoVal and median RT across participants. If fast responders systematically score higher on infoVal, something about the measurement is off.

Per-producer infoVal z-scores are computed via the package-native infoval() pipeline; both Brief-RC and 2IFC paradigms are supported.

Examples

if (FALSE) { # \dontrun{
sim <- simulate_2ifc_data(n_per_condition = 10, n_trials = 60, seed = 1)
check_rt_infoval_consistency(sim$data, method = "2ifc",
                             rdata = sim$rdata_path, col_rt = "rt")
} # }