MerQuR+Statistics Glossary

Statistics Glossary → Heritability (h²)

Heritability (h²)concept

The proportion of the total (phenotypic) variance in a trait that is due to genetic differences.

Heritability is computed from variance components: H² = σ²_G / σ²_P, the ratio of genetic to phenotypic variance. Broad-sense heritability (H²) puts all genetic variance in the numerator, while narrow-sense heritability (h²) uses only the additive genetic variance; in breeding it is narrow-sense heritability that predicts the response to selection. In plant breeding trials the genotype, environment, genotype x environment and error variances are estimated with a mixed model, and on a genotype-mean basis formulas such as H² = σ²_G / (σ²_G + σ²_GxE / e + σ²_error / (r·e)) are used (e environments, r replicates).

Heritability is specific to a population and an environment: the same trait can show different heritability in another population or in a more variable environment. It does not mean 'what percentage of the trait is genetic'; it describes the source of differences between individuals, not within an individual. Because variance-component estimates can be very uncertain with few genotypes or replicates, heritability should be judged together with a confidence interval.

An everyday example

A breeder grows 50 wheat varieties in three fields, with three replicates in each. Part of the total variation in grain yield comes from the varieties' genetics, part from field conditions and part from chance. If heritability comes out at 0.70, most of the differences between variety means are genetic, and selecting the highest-yielding varieties will also raise yields in future generations.

In MerQur

The 'REML section and h²' output of the Nested Mixed Model, given in every run, computes individual and family-mean heritability from the variance components as h²_i = 4σ²_F / (σ²_F + σ²_RF + σ²_e) (half-sib assumption). Variance Components Analysis estimates the variance components needed in heritability studies by REML, ML, ANOVA Type I or MINQUE.

Analyses where this concept appears

Nested Mixed Model (Nested LMM, Multi-Level Hierarchical Mixed Model)Variance Components Analysis (VARCOMP)

Related concepts

Variance components