Check response missingness against observed predictors — check

check_missingness() screens whether response missingness is associated with observed variables by fitting a logistic model for the missing-response indicator. This is a practical MCAR/MAR review: associations with observed predictors are compatible with missing at random after conditioning on those predictors, while missing not at random cannot be confirmed or ruled out from observed data alone.

Usage

check_missingness(
  data,
  response_var,
  predictors = NULL,
  time_var = NULL,
  subject_var = NULL,
  alpha = 0.05,
  max_factor_levels = 20
)

Arguments

data: Data frame containing the response and candidate predictors.
response_var: Response column name.
predictors: Optional predictor column names. Defaults to observed columns other than the response, subject identifier, and stored simulation truth columns.
time_var, subject_var: Optional time and subject identifier column names. time_var is included as a predictor when present; subject_var is excluded by default.
alpha: Significance threshold used to flag predictor associations.
max_factor_levels: Maximum number of levels allowed for default factor predictors.

Value

An object of class gamlss_longitudinal_missingness_check.