Worked example: native irregular sampling

This workflow separates representation from projection.

1. Keep the original grids

irregular = from_irregular_long_dataframe_native(
    samples,
    curve_columns=["participant_id", "trial_id"],
    time_column="time_s",
    value_columns=["x", "y"],
    coordinate_system="normalized",
    time_unit="s",
)

2. Audit sampling

irregular_sampling_summary(irregular)

A large maximum interval may reflect a blink, dropped packets, tracker loss, or another acquisition issue. The package does not guess which.

3. Define the common interval

grid = make_common_grid(
    irregular,
    n_time=121,
    domain="overlap",
)

4. Project with a maximum permitted gap

gaze = resample_irregular_to_grid(
    irregular,
    grid,
    method="linear",
    max_gap=0.10,
)

If unresolved missing values remain, grid FPCA will refuse them. That failure is intentional: the analyst must decide how the study treats those trials.