Error in regress_data when a trial is completely NA

[TommasoGhilardi]

If regress_data is used on a dataset in which a trial is completely NA an error is given:

Error in mutate(): ℹ In argument: pupil1newkk = .predict_right(L_P, R_P). ℹ In group 1: Subject = "Subject_1", TrialN = 1, Event = "Circle". Caused by error in lm.fit(): ! 0 (non-NA) cases

I beleive this was already reported in #19 but i don't believe was solved.

Suggested Fix:

Modify the internal .predict_right function used within mutate to include a check for cases where both columns are entirely NA. If this condition is met, the function should return a vector of NA with the same length as the input.

.predict_right <- function(x, y) {
  if (all(is.na(x)) & all(is.na(y))) {
    return(rep(NA, length(x))) # Return a vector of NA if both columns are entirely NA
  }
  pupilz <- predict(lm(x ~ y, na.action = na.exclude))
  return(pupilz)
}