Conventional tests used for behavioral phenotyping frequently have difficulties differentiating certain genotypes and replicating these differences in other laboratories. We argue that some inter-laboratory differences are unavoidable, and that this should be reflected in the statistical analysis of results from such experiments by using mixed models, rather than the conventional fixed factors ANOVA. Further limitations should be imposed on the interpretation of analytical results by controlling the False Discovery Rate (FDR) - a new approach to the multiple comparisons problem.

We demonstrate this double-headed strategy in the multiple endpoints analysis of open field behavior of inbred mouse strains over three laboratories using SEE (Software for the Exploration of Exploration). Strain differences are evident, even though taking the mixed models approach and addressing multiplicity present higher demands on the discriminatory power. These higher demands greatly enhance the replicability of the strain differences discovered.

© 2002 Noldus Information Technology bv