Principal components analysis (PCA) is used to prepare illness data for analysis of growth and morbidity in 118 Taiwanese infants. Twelve correlated illness variables are reduced by PCA to a few, uncorrelated components summarizing the essential information of the original set. Analyses of illness variables are performed for frequency and duration measures in the first and second semesters of life for the combined sexes and for each sex. Although minor differences are apparent in the number and order of components generated for each measure by semester and sex, the pattern of loadings of raw variables on components is consistent across all categories. The components are readily interpretable in terms of the loadings as biologically coherent disease entities. Two major components stand out: diarrheal and lower respiratory illnesses. Minor components are also defined for acute gastroenteritis, tonsillitis, infant pneumonia, and upper respiratory illness. Uncorrelated illness components can be used as independent variables in regression analyses of growth without risking distortion caused by multicollinearity. This approach allows more accurate and sensitive analyses and may serve as a model for other forms of research on the functional consequences of disease.
Baumgartner, Richard N. and Mueller, William H.
"Multivariate Analyses of Illness Data for Use in Studies on The Relationship of Physical Growth and Morbidity,"
1, Article 9.
Available at: https://digitalcommons.wayne.edu/humbiol/vol56/iss1/9