Skip to contents

Report what the follow-up confounder fill changed

Source: R/tteenrollment_fill.R
tteenrollment_fill_summary.Rd

The function subtracts the aggregates of the filled panel from the aggregates of the raw panel. The paper reports the result.

Usage

tteenrollment_fill_summary(raw, imp)

Arguments

raw

A TTEEnrollment object before the fill, or its aggregates table from tteenrollment_fill_aggregates().

imp

A TTEEnrollment object after the fill, or its aggregates table.

Value

A data.table, one row per confounder, with columns:

confounder

The confounder name.

rows_n

Rows in the panel.

rows_filled_n

Rows the fill gave a value.

trials_n

Person-trials in the panel.

trials_filled_n

Person-trials the fill left with no NA.

trials_entry_imputed_n

Person-trials whose snapshot value was NA in the raw panel.

rows_from_imputed_entry_n

Rows the imputed snapshot value reaches.

Details

Both arguments MUST describe the same panel. The function compares the key_digest of the two tables and stops when they differ. It also stops when the two confounder sets differ. It matches the rows of imp to the rows of raw by confounder name. The two tables MAY therefore list the same confounders in a different order.

An impute_fn passed to $s1_generate_enrollments_and_ipw() MUST NOT change the row set. s1e takes the raw aggregates before it calls the callback, and this function after. A callback that adds a row or drops a row moves the key_digest. A change to an id_var, tstart_var or tstop_var value moves it too, and s1e stops here. A reorder is safe, because the digest sorts the three key columns first.

See also

tteenrollment_fill_aggregates() for the input tables.

Other tte_methods: tteenrollment_combined_combine(), tteenrollment_fill_aggregates(), tteenrollment_fill_followup_confounders(), tteenrollment_irr_combine(), tteenrollment_rates_combine(), tteenrollment_rbind()

Examples

d <- data.table::data.table(
  enrollment_person_trial_id = rep(c("a", "b"), each = 3L),
  tstart = rep(c(0L, 4L, 8L), 2L),
  tstop = rep(c(4L, 8L, 12L), 2L),
  exposed = rep(c(TRUE, FALSE), each = 3L),
  event = 0L,
  income = c(NA, "mid", NA, NA, NA, "high"),
  .tte_entry__income = rep(c("low", "mid"), each = 3L)
)
design <- TTEDesign$new(
  treatment_var = "exposed",
  outcome_vars = "event",
  confounder_vars = "income",
  follow_up_time = 12L
)
trial <- TTEEnrollment$new(d, design)
raw <- tteenrollment_fill_aggregates(trial)
tteenrollment_fill_followup_confounders(trial)
tteenrollment_fill_summary(raw, trial)
#>    confounder rows_n rows_filled_n trials_n trials_filled_n
#>        <char>  <int>         <int>    <int>           <int>
#> 1:     income      6             4        2               2
#>    trials_entry_imputed_n rows_from_imputed_entry_n
#>                     <int>                     <int>
#> 1:                      0                         0