An evaluation of bias in k-factor analysis.

Randomization and simulation are used to detect bias in k-factor analysis. In nine previously published data sets there is strong evidence of bias. This may result from either non-independence of observations or the arithmetic relationship used to estimate k-factors, which can generate "spurious correlations". Randomization can be used to test for density dependence without bias. This procedure confirms the existence of densitydependent effects in 8 of the 9 populations and 11 of the 16 k-factors previously thought to have density-dependent effects.