Audit split-half reliability of fixation or AOI metrics

Computes odd-even or random split-half reliability for common fixation and AOI-derived metrics. The audit returns the split-half correlation and the Spearman-Brown corrected reliability estimate.

Usage

audit_gazepoint_fixation_reliability(
  data,
  subject_col,
  trial_col,
  metric = c("fixation_count", "mean_fixation_duration", "total_fixation_duration",
    "aoi_dwell_prop", "transition_count", "entropy_score"),
  duration_col = NULL,
  aoi_col = NULL,
  target_aoi = NULL,
  time_col = NULL,
  group_cols = NULL,
  min_trials = 4,
  split_method = c("odd_even", "random"),
  seed = NULL,
  correlation_method = c("pearson", "spearman")
)

Arguments

data

A fixation-level or AOI-event-level data frame.

subject_col

Character scalar. Subject identifier column.

trial_col

Character scalar. Trial identifier column.

metric

Metric to audit. Supported values are "fixation_count", "mean_fixation_duration", "total_fixation_duration", "aoi_dwell_prop", "transition_count", and "entropy_score".

duration_col

Optional duration column. Required for duration metrics. Optional for "aoi_dwell_prop"; if omitted, row proportions are used.

aoi_col

Optional AOI column. Required for AOI, transition, and entropy metrics.

target_aoi

Optional AOI label required when metric = "aoi_dwell_prop".

time_col

Optional time column used to order AOI sequences.

group_cols

Optional grouping columns. Reliability is computed separately within each group.

min_trials

Minimum number of trials per subject required for inclusion.

split_method

"odd_even" or "random".

seed

Optional random seed used when split_method = "random".

correlation_method

Correlation method passed to stats::cor().

Value

A data frame containing split-half reliability diagnostics.

Examples

dat <- expand.grid(
  subject = paste0("S", 1:6),
  trial = paste0("T", 1:4),
  KEEP.OUT.ATTRS = FALSE
)
dat$duration <- rep(seq_len(6), each = 4) + rep(c(0, 0.1, 0, 0.1), 6)

audit_gazepoint_fixation_reliability(
  dat,
  subject_col = "subject",
  trial_col = "trial",
  metric = "total_fixation_duration",
  duration_col = "duration"
)
#>                    metric split_method correlation_method split_half_r
#> 1 total_fixation_duration     odd_even            pearson    0.5055624
#>   spearman_brown n_subjects_total n_subjects_used n_trials min_trials
#> 1      0.6715928                6               6       24          4
#>   reliability_status reliability_warning
#> 1                 ok