A non-parametric approach to estimate the total deviation index for non-normal data

Area of Science:

• Biostatistics
• Quantitative Methods
• Medical Informatics

Background:

• Concordance indices assess agreement between measurement methods.
• Total Deviation Index (TDI) quantifies differences in readings from the same subject across methods.
• Existing TDI methods often assume normal data and linearity, limiting their applicability.

Purpose of the Study:

• Introduce a novel non-parametric methodology for estimating and inferring the TDI.
• Develop a flexible approach applicable to any quantitative data type.
• Address limitations of current TDI estimation methods.

Main Methods:

• Developed a non-parametric statistical methodology for TDI estimation and inference.
• Applied the new method to two real-world case examples with non-normal data (skewed and count data).
• Conducted a simulation study to compare the performance of the new approach against established methods.

Main Results:

• The non-parametric methodology effectively estimates TDI for non-normal data.
• Demonstrated the applicability of the new approach in practical scenarios.
• Simulation results indicate competitive or superior performance compared to traditional methods in specific contexts.

Conclusions:

• The proposed non-parametric approach offers a robust and versatile alternative for TDI estimation.
• This methodology expands the utility of concordance measures to a wider range of data distributions.
• Facilitates more accurate agreement assessment in diverse scientific and clinical applications.