Sampling distributions of p(pos) and p(neg)

Area of Science:

• Medical Statistics
• Clinical Measurement
• Biostatistics

Background:

• Assessing the reliability of dichotomous variables (e.g., medical diagnoses, treatment recommendations) is crucial in clinical practice.
• A gold standard for comparison is often unavailable, necessitating alternative methods for evaluating measurement properties.
• Proportions of positive agreement (ppos) and negative agreement (pneg) are key metrics for evaluating inter-rater reliability.

Purpose of the Study:

• To develop a statistical framework for assessing the clinimetric properties of dichotomous variables in the absence of a gold standard.
• To provide methods for calculating confidence intervals and conducting hypothesis tests for ppos and pneg.
• To enhance the reliability assessment of clinical judgments and diagnostic tools.

Main Methods:

• Utilized a conditional binomial distribution to derive the theoretical sampling distributions of ppos and pneg.
• Developed statistical methodologies for constructing confidence intervals around ppos and pneg estimates.
• Formulated hypothesis testing procedures to evaluate the significance of observed agreement proportions.

Main Results:

• The derived sampling distributions provide a basis for statistical inference on ppos and pneg.
• Confidence intervals can be reliably constructed for both positive and negative agreement proportions.
• Hypothesis tests allow for formal statistical comparisons of agreement metrics.

Conclusions:

• Reporting ppos and pneg is essential for evaluating clinimetric properties when a gold standard is absent.
• The derived sampling distributions and associated statistical tests offer robust methods for reliability assessment.
• This approach improves the scientific rigor of evaluating dichotomous clinical variables and decision-making.