@article {Leask:2003:1357-650X:307,
title = "Principal curve analysis avoids assumptions of dependence between measures of hand skill",
journal = "Laterality: Asymmetries of Body, Brain, and Cognition",
parent_itemid = "infobike://routledg/plat",
publishercode ="routledg",
year = "2003",
volume = "8",
number = "4",
publication date ="2003-01-01T00:00:00",
pages = "307-316",
itemtype = "ARTICLE",
issn = "1357-650X",
eissn = "1464-0678",
url = "https://www.ingentaconnect.com/content/routledg/plat/2003/00000008/00000004/art00002",
doi = "doi:10.1080/13576500412331325352",
author = "Leask, Stuart",
abstract = "The relationship between laterality and performance is unclear, and using laterality indices to explore this relationship can be misleading. An alternative approach is to consider laterality as a bivariate construct. A scatterplot of measures on each side will allow the distribution
of values in a population to be visualised. However, determining where mean values lie in a population is problematic. Even nonparametric regression techniques such as lowess make assumptions about one variable being independent and the other dependent. Such assumptions are arbitrary in the
case of measures by side. Assumptions of dependence are avoided using principal curve analysis. The results of two nonparametric fitting techniques (lowess regression and a principal curve analysis) are compared using hand skill data on 12,782 11-year-olds from a UK national birth cohort.
The lowess regression is misleading. The principal curve analysis shows that absolute right-left performance differences increase with hand skill to a maximum about the point of average hand skill, and the difference is then constant up to extremes of hand skill.",
}