Methods for analysis of unbalanced, longitudinal, growth data.

Area of Science:

• Biometrics
• Human Growth and Development
• Statistical Modeling

Background:

• Traditional growth analysis often requires assumptions about functional patterns and struggles with irregularly spaced, individual-specific data.
• Evaluating statistical significance and individual variations in growth trajectories presents analytical challenges.

Purpose of the Study:

• To present a novel, assumption-free approach for analyzing growth patterns from individual-specific, irregularly spaced data.
• To enable robust estimation of growth curves, statistical evaluation of curve features, and assessment of inter-group differences.
• To facilitate the study of relationships between relative size-for-age and other individual characteristics.

Main Methods:

• Utilized LOWESS (Locally Weighted Scatterplot Smoothing), a nonparametric method, for estimating growth curves within subject groups.
• Employed the jackknife, a sample reuse technique, for evaluating the statistical significance of growth curve features and group differences.
• Calculated residuals relative to estimated curves to assess individual consistency and score relative size-for-age.

Main Results:

• Developed a method for estimating growth curves without pre-defined functional pattern assumptions.
• Demonstrated the utility of the jackknife for assessing statistical significance in growth curve analysis and group comparisons.
• Enabled the scoring of individual relative size-for-age based on residual consistency, allowing for correlational studies with other traits.

Conclusions:

• The proposed LOWESS and jackknife-based approach offers a flexible and powerful tool for analyzing complex growth data.
• This methodology can be applied to various growth-related research questions and potentially to other biological variables beyond body size.
• The approach is adaptable for testing specific hypotheses, such as sex differences in growth patterns.