The effect of repeated measurements and within-individual variance on the estimation of heritability: a simulation study

Abstract

The estimation of heritability is a common practice in the field of ecology and evolution. Heritability of the traits is often estimated using one single measurement per individual, although many traits (especially behavioural and physiological traits) are characterized by large within-individual variance, and ideally a large number of within individual measurements can be obtained. Importantly, the effect of the within-individual variance and the rate at which this variance is sampled on the estimation of heritability has not been thoroughly tested. We fill this gap of knowledge with a simulation study, and assess the effect of within- and between-individual sample size, and the true value of the variance components on the estimation of heritability. In line with previous studies we found that the accuracy and precision of heritability estimation increased with sample size and accuracy with higher values of additive genetic variance. When the sample size was above 500 accuracy and power of heritability estimates increased in the models including repeated measurements, especially when within-individual variance was high. We thus suggest to use a sample of more than 100 individuals and to include more than two repeated measurements per individual in the models to improve estimation when investigating heritability of labile traits. Significance statement Heritability reflects the part of the trait’s phenotypic variation underlined by genetic variation. Despite the difficulties of heritability calculation (high number of individuals is needed with known relatedness), it is a widely used measure in evolutionary studies. However, not every factor potentially affecting the quality of heritability estimation is well understood. We thus investigated with a comprehensive simulation study how the number of repeated measurements per individuals and the amount of within-individual variation influence the goodness of heritability estimation. We found that although the previously described effect of the number of studied individuals was the most important, including repeated measurements also improved the reliability of the heritability estimates, especially when within-individual variation was high. Our results thus highlight the importance of including repeated measurements when investigating the heritability of highly plastic traits, such as behavioural or physiological traits.