Related Experiment Video
Updated: Jul 17, 2026
07:56
Evaluating Regional Pulmonary Deposition using Patient-Specific 3D Printed Lung Models
Published on: November 11, 2020
4.2K
A refined spirometry dataset for comparing segmented (piecewise) linear models to that of GAMLSS
1Department of Physiology and Membrane Biology, Tupper Hall, Rm 4327, 1275 Med Sciences Drive, University of California, Davis, CA 95616, United States.
Data in Brief
|December 31, 2024
Summary
Segmented Linear Regression offers a simpler alternative to complex Generalized Additive Models for Location, Scale, and Shape (GAMLSS) for creating spirometry reference equations. This method shows comparable accuracy and good agreement in classifying lung function patterns.
Area of Science:
- Pulmonary Medicine
- Biostatistics
- Medical Informatics
Background:
- Generalized Additive Models for Location, Scale, and Shape (GAMLSS) are standard for spirometric reference equations but are complex.
- Spline tables are often required for GAMLSS, increasing complexity.
- Simpler statistical models may offer viable alternatives for spirometry reference equation development.
Purpose of the Study:
- To evaluate Segmented Linear Regression as a simpler alternative to GAMLSS for spirometry reference equations.
- To compare the predictive accuracy and agreement of Segmented Linear Regression and GAMLSS models.
- To assess the classification of spirometric patterns using both modeling approaches.
Main Methods:
- Analysis of NHANES 2007-2012 spirometry data (n≈16,600) with Grade A/B quality.
- Generation of reference equations for FEV1, FVC, and FEV1/FVC using GAMLSS and Segmented Linear Regression.
- K-fold cross-validation with RMSE and correlation coefficients for accuracy comparison.
- Kappa statistic to evaluate agreement in spirometric pattern classification (obstruction, restriction, mixed).
- Inclusion of Lower Limit of Normal (LLN) using z-scores (-1.645 or -1.96).
Main Results:
- Segmented Linear Regression demonstrated comparable predictive accuracy to GAMLSS for FEV1 and FVC.
- Good agreement was observed between the two methods in classifying spirometric patterns.
- The study provides a comparative analysis of statistical modeling techniques for pulmonary function reference equations.
- Publicly available dataset (SPSS, CSV) facilitates further research.
Conclusions:
- Segmented Linear Regression is a promising, less complex alternative for developing spirometry reference equations.
- The choice of model impacts spirometric pattern classification, highlighting the need for careful consideration.
- This research contributes to the development of accessible and accurate tools for respiratory health assessment.
