Variance Partitioning and Effects of Sex and Age on Morphometric Traits in Randombred House Mice

Publisher Website
Google Scholar
Abstract
The effects of sex, age, sex by age interactions, and error were estimated from (unbalanced) two-way analyses of variance for 18 morphometric traits in 1-, 3-, and 5-month-old randombred house mice. Variance component estimates generally were higher for sex and age (and lower for error) than were estimates from contributions of sums of squares (SSQ) to the total SSQ, but both sets of values were congruent over all characters. Average variance component estimates due to sex, age, interaction, and error, respectively, were 6, 80, 1, and 13% for all three age groups (population A), but 33, 12, 0, and 55% for only the 3- and 5-month-old individuals (population B). The variance components showed significant heterogeneity among the characters for all categories except the age by sex interaction, obturator foramen length exhibiting the greatest sexual dimorphism, femur length the greatest age contributions. The variance component technique was considered preferable to the SSQ approach to variance partitioning, primarily because the SSQ contributions vary with different cell numbers and/or sizes. Estimation of fixed effects such as sex should be done by treating them as random in a variance components analysis rather than by subtraction of all other effects from the total variance.

This publication has 1 reference indexed in Scilit: