Correlating principal crickets
It is wonderful to see ecologists asking structural questions about personality using measures of many behaviors but I am a bit puzzled by the statistical analysis in this paper by Wilson et al on house crickets:
We tested whether laboratory-reared male and female European house crickets, Acheta domesticus, exhibited behavioral syndromes by quantifying individual differences in activity, exploration, mate attraction, aggressiveness, and antipredator behavior. To our knowledge, our study is the first to consider such a breadth of behavioral traits in one organism using the syndrome framework.Their goal is to identity a boldness syndrome and they start sensibly by measuring a slew of behaviors such as time spent walking or climbing and latency to emerge in different contexts. Strangely, these behaviors are pre-placed into categories (defined as "behavioral contexts" with 2 to 4 behaviors each). The behaviors in each context were subjected to a principal components analysis and then the first principal components were correlated with each other to look for a syndrome. My first question is: what software were they using that didn't throw up an error when doing a PCA on only two variables (in general you need at least 3 variables, otherwise you are just getting the correlation between the two variables). Second, why not instead do a PCA on all the raw variables? This would be another way to look for behavioral suites across contexts. There may be latent variables here that span the presupposed behavioral context. Wilson et al. Behavioral correlations across activity, mating, exploration, aggression, and antipredator contexts in the European house cricket, Acheta domesticus. Behav Ecol Sociobiol (2009) doi:10.1007/s00265-009-0888-1 photo cc-by alles-schlumpf
