Process longitudinal dyadic data in wide format with censoring by missing exposure and silent dummy-coding
Source:R/margot_process_longitudinal_data_wider.R
margot_process_longitudinal_data_wider.Rdthis function processes longitudinal data (wide format) across multiple waves, handling dyadic censoring and optional censoring when the exposure is missing.
if `censor_if_missing_exposure = TRUE`, any record with a missing exposure at wave t+1 is considered lost at wave t and all subsequent wave data are set to `NA`. dyadic logic ensures that if any member of a dyad is lost, the entire dyad is censored.
Usage
margot_process_longitudinal_data_wider(
df_wide,
relationship_id = "NULL",
ordinal_columns = NULL,
continuous_columns_keep = NULL,
exposure_vars = NULL,
scale_exposure = FALSE,
scale_continuous = TRUE,
censor_if_missing_exposure = TRUE,
not_lost_in_following_wave = "not_lost_following_wave",
lost_in_following_wave = NULL,
remove_selected_columns = TRUE,
time_point_prefixes = NULL,
time_point_regex = NULL,
save_observed_y = FALSE
)Arguments
- df_wide
a wide-format `data.frame` containing time-prefixed columns (e.g., `t0_x`).
- relationship_id
column name identifying dyads; if present, dyadic censoring is applied. default: `"NULL"`.
- ordinal_columns
character vector of ordinal column bases to dummy-encode after processing.
- continuous_columns_keep
character vector of continuous column names to retain without scaling.
- exposure_vars
character vector of exposure base names (without time prefixes).
- scale_exposure
logical; if `TRUE`, scales exposure variables. default: `FALSE`.
- scale_continuous
logical; if `TRUE`, scales continuous variables. default: `TRUE`.
- censor_if_missing_exposure
logical; if `TRUE`, missing exposures at next wave cause censoring. default: `TRUE`.
- not_lost_in_following_wave
suffix for the not-lost indicator. default: `"not_lost_following_wave"`.
- lost_in_following_wave
suffix for the lost indicator; if `NULL`, no lost indicator is added.
- remove_selected_columns
logical; if `TRUE`, removes original columns after dummy encoding. default: `TRUE`.
- time_point_prefixes
optional vector of time-point prefixes (e.g., `c("t0","t1")`). if `NULL`, inferred from data.
- time_point_regex
regex pattern to identify time-point prefixes. used if `time_point_prefixes` is `NULL`.
- save_observed_y
logical; if `TRUE`, retains observed outcomes in final wave even if censored. default: `FALSE`.