Analyzing test batteries in animal models of psychopathology with multivariate analysis of variance (MANOVA): One...

One concern regarding animal models of psychopathology is unclear external validity. One way to establish external validity is to examine measures representing separate facets of the pathology with a battery of tests in the same cohort of animals. Additionally, utilizing the same animals in a battery of tests can help to reduce the number of animals in research. However, issues had been raised regarding the analysis of data coming from batteries and the standard practice is to analyze each test separately. This approach introduces two problems: (1) the analysis answers the question regarding separate tests but not regarding the general effect; (2) there is no correction for multiple comparisons. One way to overcome these challenges is to use transformations to Z-scores. We suggest an additional approach, analyzing test batteries with multivariate analysis of variance (MANOVA).

To compare the outcomes of Z-score analysis and MANOVA we re-analyzed two published studies where data were initially analyzed separately for each test. Additionally, we computed effect sizes.

The first study tested interaction between sex and lithium in a battery of manic-like behaviors, the second study tested asenapine in a battery of anxiety-like behaviors. For the first study, the MANOVA analysis indicated no effects of sex and a significant antimanic-like effect of lithium and for the second study, the MANOVA indicated a significant anxiolytic effect of asenapine. Z-score analysis resulted in a significant general antimanic-like effect in the lithium study but failed to demonstrate the anxiolytic effects of asenapine in the second study.

It is possible to suggest that MANOVA is an appropriate way to analyze data from test batteries and that its use, when appropriate, can increase the validity, predictability and reproducibility of results.