An evaluation of the variance components approach: type I error, power and size of the estimated effect.

In order to evaluate the statistical properties of the variance components method implemented in SOLAR and GENEHUNTER2, the type I error rate, power and estimated size of modelled effects were determined using computer simulation. Results suggested that the type I error rate was quite conservative when the variance was completely due to random effects. However, when either a polygenic effect or an unlinked single locus component effect was included in the generating models the type I error rate was closer to the nominal level. The size of the polygenic or single locus effect appeared to influence the type I error rate. Results suggested that the variance components method underestimated the variance attributed to a single locus effect and overestimated the variance attributed to a polygenic effect in the models considered regardless of the source of variation.