Impute missing pupil samples with a lightweight Gaussian-process smoother
Source:R/experimental_bayesian_bridges.R
impute_gazepoint_pupil_gp.RdPerforms within-subject/trial Gaussian-process interpolation using a squared exponential kernel. This helper is intended for short missing segments after blink detection, not for reconstructing long unusable trials.
Usage
impute_gazepoint_pupil_gp(
data,
pupil,
time,
subject = NULL,
trial = NULL,
length_scale = NULL,
noise = 1e-04,
max_train = 300,
output = "pupil_gp_imputed",
flag = "pupil_was_gp_imputed"
)Arguments
- data
A data frame.
- pupil
Pupil column.
- time
Time column.
- subject
Optional subject column.
- trial
Optional trial column.
- length_scale
Kernel length scale in the same unit as
time.- noise
Observation noise variance.
- max_train
Maximum number of observed samples used per sequence.
- output
Name of the imputed output column.
- flag
Name of the logical imputation flag column.