Passing-Bablok regression and measurement error

This project focused on agreement between two ways of measuring infant breast bud diameter: ultrasound-based measurements and bead-based measurements. I used Passing-Bablok regression, a method designed for method-comparison settings, to better understand how closely the two approaches aligned across sex and study visit.

The work reflects my interest in measurement error and in the practical side of statistical analysis. Before interpreting a scientific result, it is often important to understand how the measurements themselves compare and where disagreement may enter the analysis.

Methods

• Compared ultrasound-based and bead-based geometric mean breast bud diameter
• Used Passing-Bablok regression to estimate intercepts and slopes with confidence intervals
• Stratified analyses by infant sex and study visit
• Interpreted departures from the line of identity as evidence of disagreement between methods

Key findings

• The relationship between bead-based and ultrasound-based measurements varied across visits and by sex
• Estimated slopes were often below 1 at earlier visits, suggesting the two methods did not always scale identically
• Sample size changed across visits because bead-based measurements were unavailable for infants without a palpable bud and some visits had additional missingness

Personal + practical importance

This project strengthened my understanding of how measurement decisions shape downstream analysis. It also gave me experience presenting statistical comparison results in a way that is directly useful for substantive researchers working with developing health measures.