Statistical models for assessing agreement for quantitative data with heterogeneous random raters and replicate...

Area of Science:

• Biostatistics
• Statistical Modeling
• Method Comparison Studies

Background:

• Traditional agreement assessment relies on Bland-Altman plots or limits of agreement for paired methods.
• Situations arise where measurement methods are a random sample from a larger population of potential methods.

Purpose of the Study:

• To extend linear mixed effects models for situations involving random measurement methods.
• To quantify variation between random methods and incorporate this into clinical performance generalization.
• To allow for individual rater precision and linked replicates within the model.

Main Methods:

• Extension of linear mixed effects models to accommodate random measurement methods.
• Incorporation of individual rater precision and linked replicates.
• Implementation using the MethComp package in R.

Main Results:

• Demonstration of estimating limits of agreement for two datasets using the proposed model.
• Application to spatial perception and consumer preference data.
• Quantification of inter-method variation as an additional source of error.

Conclusions:

• The extended model provides a robust framework for method comparison when methods are considered random.
• This approach offers improved generalization of clinical performance by accounting for method variability.
• The MethComp package facilitates the application of these advanced statistical techniques.