Test Data for groomecompare Function

Format

groomecompare_test: Standard dataset with 150 observations and 7 variables:

patient_id

Character. Patient identifier (PT001-PT150)

time

Numeric. Survival time in months (exponential distribution, mean ~20)

event

Factor. Event indicator (0 = censored, 1 = death/event). Event rate ~60\ ageNumeric. Patient age in years (40-80) sexFactor. Patient sex (Male, Female) ypTNMOrdered factor. Post-neoadjuvant pathological staging (I, II, III, IV) RPAOrdered factor. Recursive partitioning risk groups (Low, Intermediate, High Risk)

groomecompare_small: Small sample dataset with 60 observations and 7 variables: groomecompare_large: Large dataset with 300 observations and 8 variables: Special test datasets:

groomecompare_unbalanced

120 observations with unbalanced staging systems (5 vs 2 groups)

groomecompare_tied

80 observations with many tied survival times

groomecompare_identical

100 observations with identical staging systems (negative control)

groomecompare_clear_winner

150 observations where one system is clearly superior

groomecompare_edge_truefalse

40 observations with event coded as TRUE/FALSE

groomecompare_edge_12

40 observations with event coded as 1/2

Generated synthetically using data-raw/groomecompare_test_data.R. Seed: 12345. Generation date: 2026-01-31. Synthetic datasets for testing and demonstrating the groomecompare function (Groome Staging System Comparison for Survival Data). These datasets were generated using a seeded random number generator to produce realistic survival data for testing Groome staging system comparison with the following characteristics:

• Survival times follow exponential distribution

• Event rates are clinically realistic (55-65\

• Prognostic correlations built in (advanced stage → shorter survival)

• Both systems have predictive value but differ in discrimination

• Sufficient events per variable (EPV > 10) for Cox models

The Groome method (Groome et al., 2001) compares staging systems using four criteria:

1. Hazard Consistency: Monotonicity of hazard ratios across stages

2. Hazard Discrimination: Range/spread of hazard ratios between stages

3. Sample Balance: Distribution of patients across staging groups

4. Outcome Prediction: C-index/concordance for outcome prediction

The data generation process ensures:

• Non-negative survival times

• Proper factor level ordering (ordinal staging variables)

• Realistic clinical distributions

• Differential performance between staging systems

• Sufficient sample sizes for comparison

File Formats Each dataset is available in multiple formats:

• RDA: Native R format (use data())

• CSV: Comma-separated values (data/nonrda/)

• XLSX: Excel format (data/nonrda/)

• OMV: jamovi native format (data/nonrda/)

Multi-sheet workbook groomecompare_all_scenarios.xlsx contains all test scenarios. Usage Examples See vignette("groomecompare-examples") for comprehensive examples. Basic usage:

data(groomecompare_test)
library(ClinicoPath)

# Standard Groome comparison
groomecompare(
  data = groomecompare_test,
  time = "time",
  event = "event",
  stage1 = "ypTNM",
  stage2 = "RPA",
  stage1name = "ypTNM Staging",
  stage2name = "RPA Classification"
)

# With bootstrap validation
groomecompare(
  data = groomecompare_test,
  time = "time",
  event = "event",
  stage1 = "ypTNM",
  stage2 = "RPA",
  bootstrap = TRUE,
  nboot = 100,
  seed = 12345
)

# Test different event coding
data(groomecompare_edge_truefalse)
groomecompare(
  data = groomecompare_edge_truefalse,
  time = "time",
  event = "event_tf",
  stage1 = "ypTNM",
  stage2 = "RPA",
  eventValue = "TRUE"
)

Testing Scenarios The datasets support testing of:

1. Standard comparison: Use groomecompare_test with two staging systems

2. Small samples: Use groomecompare_small, test sample size warnings

3. Complex systems: Use groomecompare_large with detailed AJCC8 vs RPA5

4. Unbalanced groups: Use groomecompare_unbalanced (5 groups vs 2 groups)

5. Tied times: Use groomecompare_tied to test tie handling

6. Identical systems: Use groomecompare_identical (negative control, all metrics ~0.5)

7. Clear winner: Use groomecompare_clear_winner where one system dominates

8. Event coding: Test TRUE/FALSE and 1/2 coding schemes

9. All visualizations: Test radar plots, bar plots, Kaplan-Meier curves

10. Bootstrap validation: Test with bootstrap=TRUE, nboot=100-500

Validation All datasets have been validated for:

• Non-negative survival times

• Appropriate event rates (55-65\

• Stage-survival correlation (advanced stage → worse prognosis)

• Sufficient EPV (events per variable > 10) for Cox models

• Realistic clinical distributions

• Proper factor level ordering for ordinal staging

• Differential performance between systems (for comparison testing)

Groome Criteria The four Groome criteria used for staging system comparison:

• Hazard Consistency (Rank 1-2): Monotonicity of hazard ratios

• Hazard Discrimination (Rank 1-2): Spread of hazard ratios

• Sample Balance (Rank 1-2): Distribution across stages

• Outcome Prediction (Rank 1-2): C-index comparison

• Overall Rank: Sum of individual ranks (lower is better)

# Load standard test data data(groomecompare_test) # Examine structure str(groomecompare_test) # Summary statistics summary(groomecompare_test) # Check event rate table(groomecompare_test$event) prop.table(table(groomecompare_test$event)) # Check staging distributions table(groomecompare_test$ypTNM) table(groomecompare_test$RPA) # Cross-tabulation of staging systems table(groomecompare_test$ypTNM, groomecompare_test$RPA) # Basic Groome comparison Groome PA, Schulze K, Boysen M, Hall S, Mackillop WJ. (2001). A comparison of published head and neck stage groupings in carcinomas of the oral cavity. Head Neck, 23(8):613-624. doi:10.1002/hed.1089 Balci S, Altinay S. (2025). Comparison of prognostic staging systems in gastrointestinal neuroendocrine tumors using Groome method. Turk Patoloji Derg, 41(1):1-10. doi:10.5146/tjpath.2023.01590 groomecompare for the main analysis function vignette("groomecompare-examples") for comprehensive usage examples rpasurvival_test_data for RPA survival analysis test data datasets