Wilson A.J. (2008) Why h2 does not always equal VA/VP? Journal of Evolutionary Biology. Online early. doi:10.1111/j.1420-9101.2008.01500.x
Alastair Wilson makes an important point about heritability estimates. The point is especially relevant for us looking for trends and trying to find answers to general questions on heritability.
It is usually taken into consideration that heritability estimates are population and trait specific, but too often ignored that the estimates of variance components that make up heritability are highly dependent on the model used for their estimation. Wilson demonstrates this with a simulation, showing how five-fold differences in heritability estimates can be achieved from the same data set by choosing different fixed effects for the model. Wilson does not request abandoning the current practices for heritability estimation, but suggests that the choice of model should be considered as a source of serious bias when interpreting the results of quantitative genetic studies. After all, the phenotypic variance used in heritability calculations is that which remains after correcting the variance with fixed effects of choice. Wilson also suggests that a simple estimate of phenotypic variance observed in the raw data should be provided in studies published in the future to allow the readers to calculate standardized heritability estimates, if they so will.