Psychometric Process-Data Models

In eyeprocess: Harmonize Eye-Tracking, Pupillometry, Biometrics, and Psychometric Process Data

knitr::opts_chunk$set(collapse = TRUE, comment = "#>", eval = FALSE)

The package prepares linked person-item-trial data and delegates mature IRT estimation to optional engines where appropriate.

Response and response-time matrices

y  <- response_matrix(x)
rt <- response_time_matrix(x, log_transform = TRUE)
aligned <- align_response_matrices(y, rt)

Conventional and explanatory IRT

fit_mirt <- fit_irt(x, engine = "mirt", model = 1, itemtype = "2PL")
fit_tam  <- fit_irt(x, engine = "TAM")

fit_explanatory <- fit_explanatory_irt(
  x,
  score ~ dwell_time + first_fixation_latency + pupil_auc,
  engine = "lme4"
)

Accuracy and response time

fit_rt <- fit_accuracy_rt(x, engine = "LNIRT")

LNIRT receives aligned response matrices and log response times. A two-stage fallback is available for transparent exploratory work, but it is not treated as equivalent to a joint latent model.

Process-informed and experimental models

spec <- process_irt_spec(
  response = "score",
  gaze_features = c("dwell_time", "first_fixation_latency"),
  pupil_features = c("pupil_auc"),
  response_time = "response_time"
)

fit <- fit_process_irt(x, spec, engine = "lme4")
process_irt_diagnostics(fit)

shared <- fit_shared_process_factor(
  x,
  features = c("dwell_time", "fixation_count", "pupil_auc")
)

Shared process factors are intentionally neutral labels until construct validity is established. Advanced joint and dynamic functions are marked experimental and require simulation, parameter-recovery, and empirical validation before confirmatory use.

eyeprocess documentation built on Sept. 28, 2026, 5:08 p.m.