Skip to contents

Create paired survival dates for cohort studies

Source: R/create_survival_dates.R
create_survival_dates.Rd

Generates entry and event dates with guaranteed temporal ordering (entry < event). Useful for survival analysis, cohort studies, and time-to-event modeling.

Usage

create_survival_dates(
  entry_var_row = NULL,
  entry_details_subset = NULL,
  event_var_row = NULL,
  event_details_subset = NULL,
  n = NULL,
  seed = NULL,
  df_mock = NULL,
  entry_var = NULL,
  event_var = NULL,
  entry_start = NULL,
  entry_end = NULL,
  followup_min = NULL,
  followup_max = NULL,
  length = NULL,
  event_distribution = "uniform",
  prop_censored = 0,
  prop_NA = NULL,
  prop_invalid = NULL
)

Arguments

entry_var_row

data.frame. Single row from mock_data_config for entry date variable

entry_details_subset

data.frame. Rows from mock_data_config_details for entry date

event_var_row

data.frame. Single row from mock_data_config for event date variable

event_details_subset

data.frame. Rows from mock_data_config_details for event date

n

integer. Number of observations to generate

seed

integer. Random seed for reproducibility. If NULL, uses global seed.

df_mock

data.frame. The current mock data (to check if variables already exist)

Configuration v0.1 format (LEGACY):

entry_var

character. Name for entry date variable

event_var

character. Name for event date variable

entry_start

Date. Start of entry period

entry_end

Date. End of entry period

followup_min

integer. Minimum follow-up days

followup_max

integer. Maximum follow-up days

length

integer. Number of records to generate

event_distribution

character. Distribution for time-to-event: "uniform", "gompertz", "exponential"

prop_censored

numeric. Proportion of records to censor (0-1)

prop_NA

numeric. Proportion of missing values (0-1)

prop_invalid

numeric. Optional. Proportion of temporal violations (entry > event) (0 to 1). If NULL, no invalid dates generated.

Value

data.frame with entry_date, event_date, and optionally event_status columns, or NULL if:

  • Variables already exist in df_mock

  • Missing required configuration

Details

Configuration v0.2 format (NEW):

v0.2 format (NEW):

  • Extracts date ranges from entry_details_subset and event_details_subset

  • Generates entry dates uniformly distributed

  • Calculates event dates to ensure entry < event

  • Supports garbage data via catLabel::garbage in event_details_subset

  • Supports fallback mode: reasonable defaults when details_subset is NULL

v0.1 format (LEGACY):

  • Accepts explicit date ranges and follow-up parameters

  • Supports multiple event distributions (uniform, gompertz, exponential)

  • Handles censoring, missing values, and temporal violations via parameters

The function auto-detects which format based on parameter names.

This function generates realistic survival data by:

  1. Creating entry dates uniformly distributed across entry period

  2. Generating follow-up times using specified distribution

  3. Calculating event dates (entry + follow-up)

  4. Optionally censoring events (event_status = 0)

  5. Ensuring entry_date < event_date for all records

Event distributions:

  • "uniform": Constant hazard over follow-up period

  • "gompertz": Increasing hazard (mortality increases with time)

  • "exponential": Decreasing hazard (early events more common)

Censoring: When prop_censored > 0, generates event_status column:

  • 1 = event observed

  • 0 = censored (event_date becomes censoring date)

See also

Other generators: create_cat_var(), create_con_var(), create_date_var(), create_mock_data(), create_wide_survival_data()

Examples

if (FALSE) { # \dontrun{
# v0.2 format - called by create_mock_data()
config <- read_mock_data_config("mock_data_config.csv")
details <- read_mock_data_config_details("mock_data_config_details.csv")
entry_row <- config[config$variable == "study_entry", ]
entry_details <- get_variable_details(details, variable_name = "study_entry")
event_row <- config[config$variable == "death_date", ]
event_details <- get_variable_details(details, variable_name = "death_date")
surv_data <- create_survival_dates(
  entry_var_row = entry_row,
  entry_details_subset = entry_details,
  event_var_row = event_row,
  event_details_subset = event_details,
  n = 1000,
  seed = 123
)

# v0.2 with garbage (catLabel::garbage in config_details)
# In mock_data_config_details.csv:
# variable,recEnd,catLabel,proportion
# death_date,followup_min,,
# death_date,followup_max,,
# death_date,garbage,garbage,0.03
event_details_with_garbage <- get_variable_details(details, variable_name = "death_date")
surv_data_garbage <- create_survival_dates(
  entry_var_row = entry_row,
  entry_details_subset = entry_details,
  event_var_row = event_row,
  event_details_subset = event_details_with_garbage,
  n = 1000,
  seed = 123
)

# v0.1 format (legacy) - Basic mortality study
surv_data <- create_survival_dates(
  entry_var = "study_entry",
  event_var = "death_date",
  entry_start = as.Date("2000-01-01"),
  entry_end = as.Date("2005-12-31"),
  followup_min = 365,
  followup_max = 3650,
  length = 1000,
  df_mock = data.frame(),
  event_distribution = "gompertz"
)

# v0.1 with censoring
surv_data <- create_survival_dates(
  entry_var = "cohort_entry",
  event_var = "event_date",
  entry_start = as.Date("2010-01-01"),
  entry_end = as.Date("2015-12-31"),
  followup_min = 30,
  followup_max = 1825,
  length = 500,
  df_mock = data.frame(),
  event_distribution = "exponential",
  prop_censored = 0.3
)

# v0.1 with temporal violations for validation testing
surv_data <- create_survival_dates(
  entry_var = "interview_date",
  event_var = "death_date",
  entry_start = as.Date("2015-01-01"),
  entry_end = as.Date("2016-12-31"),
  followup_min = 30,
  followup_max = 3650,
  length = 1000,
  df_mock = data.frame(),
  prop_invalid = 0.03  # 3% temporal violations
)
} # }