Skip to contents

Validate incomplete trip estimates using TOST equivalence testing

Source: R/validate-incomplete-trips.R
validate_incomplete_trips.Rd

Performs Two One-Sided Tests (TOST) to determine if incomplete trip CPUE estimates are statistically equivalent to complete trip estimates within a specified threshold. Returns validation results with recommendations for whether incomplete trips are appropriate for estimation in the given dataset.

Usage

validate_incomplete_trips(
  design,
  catch,
  effort,
  by = NULL,
  variance = "taylor",
  conf_level = 0.95,
  truncate_at = 0.5
)

Arguments

design

A creel_design object with interviews attached via add_interviews. Must include trip_status field to distinguish complete from incomplete trips.

catch

Bare column name for catch data (supports tidy evaluation)

effort

Bare column name for effort data (supports tidy evaluation)

by

Optional tidy selector for grouping variables. When provided, performs TOST for each group independently. Overall validation passes only if overall test AND all group tests pass equivalence.

variance

Character string specifying variance estimation method. Options: "taylor" (default), "bootstrap", or "jackknife". Passed to estimate_catch_rate.

conf_level

Numeric confidence level for confidence intervals (default: 0.95). Passed to estimate_catch_rate.

truncate_at

Numeric minimum trip duration (hours) for incomplete trip estimation. Default is 0.5 hours (30 minutes). Passed to estimate_catch_rate for MOR estimator.

Value

A creel_tost_validation S3 object (list) with components:

  • overall_test: List with TOST results for overall comparison (p_lower, p_upper, equivalence_passed, diff_estimate, equivalence_bounds)

  • group_tests: Data frame with per-group TOST results (only present for grouped estimation)

  • equivalence_threshold: Numeric threshold used (e.g., 0.20 for ±20\

  • passed: Logical indicating if overall AND all groups (if applicable) passed equivalence

  • recommendation: Character string with usage recommendation

  • metadata: List with complete and incomplete trip estimates, standard errors, confidence intervals, and sample sizes

Details

The function estimates CPUE separately for complete trips (using ratio-of-means) and incomplete trips (using mean-of-ratios) via estimate_catch_rate. It then performs TOST to test equivalence within the specified threshold.

Equivalence bounds are calculated as ±threshold * complete_trip_estimate. The difference variance is estimated using the delta method: Var(complete - incomplete) = Var(complete) + Var(incomplete), assuming independence between the two samples.

Sample size requirements: At least 10 complete trips AND 10 incomplete trips are required for stable variance estimation and TOST. The function errors if either sample size is insufficient.

Package Options

The package option tidycreel.equivalence_threshold controls the equivalence threshold (default: 0.20 = ±20\ bounds for equivalence as ±threshold * complete_trip_estimate. For example, with the default 20\ equivalence bounds are 1.6 to 2.4 fish/hour. Users can set a custom threshold:

options(tidycreel.equivalence_threshold = 0.15)

TOST Equivalence Testing

TOST (Two One-Sided Tests) is the statistically appropriate method for proving similarity between two estimates. Unlike traditional hypothesis testing which tests for difference, TOST tests the null hypothesis that estimates differ by MORE than the threshold. Equivalence is concluded when both one-sided tests reject the null (both p-values < 0.05).

The two tests are:

  1. H0: complete - incomplete <= -threshold vs H1: complete - incomplete > -threshold

  2. H0: complete - incomplete >= threshold vs H1: complete - incomplete < threshold

Both tests must reject (p < 0.05) for equivalence. This ensures estimates are "close enough" to be considered equivalent for practical purposes.

Grouped Validation

When by is provided, the function performs TOST for each group independently AND for the overall (ungrouped) data. The validation passes only if ALL tests pass equivalence. This conservative approach prevents overlooking group-specific bias that could be masked by overall equivalence.

See also

Other "Reporting & Diagnostics": adjust_nonresponse(), check_completeness(), compare_variance(), flag_outliers(), season_summary(), standardize_species(), summarize_by_angler_type(), summarize_by_day_type(), summarize_by_method(), summarize_by_species_sought(), summarize_by_trip_length(), summarize_cws_rates(), summarize_hws_rates(), summarize_length_freq(), summarize_refusals(), summarize_successful_parties(), summarize_trips(), summary.creel_estimates(), validate_creel_data(), validate_design(), validation_report(), write_estimates()

Examples

# Create design with both complete and incomplete trips
calendar <- data.frame(
  date = as.Date(c("2024-06-01", "2024-06-02", "2024-06-03", "2024-06-04")),
  day_type = rep(c("weekday", "weekend"), each = 2)
)
design <- creel_design(calendar, date = date, strata = day_type)

set.seed(123)
interviews <- data.frame(
  date = as.Date(rep(c("2024-06-01", "2024-06-02", "2024-06-03", "2024-06-04"), each = 25)),
  catch_total = rpois(100, lambda = 6),
  hours_fished = runif(100, min = 2, max = 4),
  trip_status = rep(c("complete", "incomplete"), each = 50),
  trip_duration = runif(100, min = 2, max = 4)
)

design_with_interviews <- add_interviews(design, interviews,
  catch = catch_total,
  effort = hours_fished,
  trip_status = trip_status,
  trip_duration = trip_duration
)
#> ℹ No `n_anglers` provided — assuming 1 angler per interview.
#> ℹ Pass `n_anglers = <column>` to use actual party sizes for angler-hour
#>   normalization.
#> Warning: 1 interview has zero catch.
#> ℹ Zero catch may be valid (skunked) or indicate missing data.
#> ℹ Added 100 interviews: 50 complete (50%), 50 incomplete (50%)

# Validate incomplete trips
result <- validate_incomplete_trips(design_with_interviews,
  catch = catch_total,
  effort = hours_fished
)
print(result)
#> 
#> ── TOST Equivalence Validation Results ─────────────────────────────────────────
#> Threshold: ±20% of complete trip estimate
#> ✔ Validation PASSED
#> 
#> Recommendation: Validation passed: Safe to use incomplete trips for CPUE
#> estimation in this dataset
#> 
#> 
#> ── Overall Test ──
#> 
#> Complete trips: n = 50, CPUE = 2.046
#> Incomplete trips: n = 50, CPUE = 1.969
#> Difference: 0.077
#> Equivalence bounds: [-0.409, 0.409]
#> TOST p-values: p_lower = 0.0049, p_upper = 0.0359
#> ✔ Overall equivalence: PASSED
#> 


# Grouped validation
result_grouped <- validate_incomplete_trips(design_with_interviews,
  catch = catch_total,
  effort = hours_fished,
  by = day_type
)
#> Warning: ! Only 0.0% of interviews are complete trips (threshold: 10%)
#> ℹ Pollock et al. recommends >=10% complete trips for valid estimation
#> ℹ Consider use_trips='diagnostic' to validate incomplete trip estimates
print(result_grouped)
#> 
#> ── TOST Equivalence Validation Results ─────────────────────────────────────────
#> Threshold: ±20% of complete trip estimate
#> ✔ Validation PASSED
#> 
#> Recommendation: Validation passed: Safe to use incomplete trips for CPUE
#> estimation in this dataset (overall and all groups passed equivalence)
#> 
#> 
#> ── Overall Test ──
#> 
#> Complete trips: n = 50, CPUE = 2.046
#> Incomplete trips: n = 50, CPUE = 1.969
#> Difference: 0.077
#> Equivalence bounds: [-0.409, 0.409]
#> TOST p-values: p_lower = 0.0049, p_upper = 0.0359
#> ✔ Overall equivalence: PASSED
#> 
#> 
#> ── Per-Group Tests ──
#> 
#> 1/1 group passed equivalence
#> 

# Custom equivalence threshold
options(tidycreel.equivalence_threshold = 0.15) # 15% threshold
result_custom <- validate_incomplete_trips(design_with_interviews,
  catch = catch_total,
  effort = hours_fished
)