valytics 0.4.1

Bug Fixes

Fixed troponin_precision dataset which was generated with constant values (zero variance) at each concentration level. The dataset now contains realistic variability following a hyperbolic CV model, enabling proper demonstration of precision_study() and precision_profile() functions.

Documentation

Updated README with precision experiments examples.

valytics 0.4.0

New Features

Precision Experiments (EP05/EP15-aligned)

• precision_study(): Comprehensive variance component analysis for precision experiments with nested experimental designs.

  • Supports multiple design types: single-site (day/run/replicate), multi-site (site/day/run/replicate), and custom nested designs
  • Automatic design detection from data structure
  • Two estimation methods: ANOVA (method of moments) and REML (via lme4)
  • Three confidence interval methods: Satterthwaite (default), Modified Large Sample (MLS), and bootstrap BCa
  • Multi-sample support for analyzing multiple concentration levels
  • Complete S3 methods: print(), summary(), plot(), autoplot()

• verify_precision(): Statistical verification of observed precision against manufacturer claims using chi-square hypothesis testing.

  • Accepts numeric vectors, precision_study objects, or data frames
  • Calculates Upper Verification Limit (UVL)
  • Provides confidence intervals for observed precision
  • Clear pass/fail determination with detailed interpretation

• precision_profile(): Models the relationship between CV and concentration for functional sensitivity estimation.

  • Hyperbolic model: CV = sqrt(a² + (b/x)²) (default)
  • Linear model: CV = a + b/x
  • Calculates functional sensitivity at user-specified CV targets
  • Optional bootstrap confidence intervals for functional sensitivity
  • Integrates with precision_study objects for seamless workflow

Precision Measures

The package now calculates and reports: - Repeatability: Within-run precision - Between-run precision: Additional variability between runs within a day - Between-day precision: Additional variability between days - Within-laboratory precision: Combined day + run + error variance - Between-site precision: Additional variability between sites (multi-site only - Reproducibility: Total precision including all variance components

Visualization

• plot.precision_study() with three plot types:
  • type = "variance": Variance component bar chart
  • type = "cv": CV profile across precision measures with CIs
  • type = "precision": Forest plot of precision estimates
• plot.precision_profile(): Publication-ready precision profile visualization
  • Fitted curve with prediction intervals
  • Functional sensitivity target lines
  • Optional logarithmic x-axis scale

New Dataset

• troponin_precision: High-sensitivity cardiac troponin I precision study data with 6 concentration levels (5-500 ng/L), designed for demonstrating precision_study() and precision_profile() workflows.

Documentation

• New vignette: “Precision Profiles and Functional Sensitivity” - demonstrates the complete workflow from raw precision data to functional sensitivity estimation with clinical interpretation.

Dependencies

• lme4 added to Suggests for REML estimation (optional)

valytics 0.3.0

Initial CRAN release.

New features

• ate_from_bv(): Calculate allowable total error (ATE) specifications from biological variation data using the Fraser-Petersen model. Supports three performance levels (optimal, desirable, minimum) and provides allowable imprecision, allowable bias, and total allowable error specifications.

• sigma_metric(): Calculate the Six Sigma metric for analytical performance assessment. Returns sigma value with interpretation category (World Class to Unacceptable) and approximate defect rates.

• ate_assessment(): Comprehensive evaluation of observed method performance against allowable total error specifications. Provides pass/fail assessment for individual components (bias, CV, total error) and overall method acceptability, integrated with sigma metric calculation.

Documentation

• New vignette: “Setting Quality Goals with Biological Variation” – explains the biological variation model for analytical performance specifications with practical examples using the new ATE functions.

valytics 0.2.0

New features

• deming_regression(): Deming regression for method comparison, accounting for measurement error in both variables. Supports known error ratio or estimation from replicates. Includes jackknife and bootstrap BCa confidence intervals.

• S3 methods for Deming regression: print(), summary(), plot(), and autoplot() (ggplot2).

Documentation

• New vignette: “Deming Regression for Method Comparison” – comprehensive guide to Deming regression theory and practical application.

• Updated vignette: “Understanding Method Comparison Statistics” – added guidance on choosing between regression methods.

valytics 0.1.0

New features

• ba_analysis(): Bland-Altman method comparison analysis with bias estimation, limits of agreement, and confidence intervals. Supports both absolute and percentage difference scaling.

• pb_regression(): Passing-Bablok regression with fast O(n log n) algorithm via the robslopes package. Includes analytical confidence intervals (Passing & Bablok 1983) and optional bootstrap BCa intervals. CUSUM test for linearity assessment with Kolmogorov-Smirnov p-value.

• S3 methods for both analyses: print(), summary(), plot(), and autoplot() (ggplot2).

• Publication-ready visualizations using ggplot2, including Bland-Altman plots, regression scatter plots with confidence bands, residual plots, and CUSUM plots for linearity assessment.

Datasets

• glucose_methods: Point-of-care glucose meter vs laboratory analyzer (n=60)

• creatinine_serum: Enzymatic vs Jaffe creatinine methods (n=80)

• troponin_cardiac: Two high-sensitivity cardiac troponin I platforms (n=50)

Documentation

• Vignette: “Method Comparison Workflow” – step-by-step analysis guide

• Vignette: “Understanding Method Comparison Statistics” – educational overview of statistical concepts for method comparison studies

mirror server hosted at Truenetwork, Russian Federation.