In order to compare how LDA performs relative to better-known and more commonly used dimension reduction method, PCA was performed on the same initial dataset as used to generate the IDs. (a) The percent variance of the behavioral data explained by each principal component (PC). (b) Correlations between scores on each PC and an abbreviated list of behavioral readouts. (c) The stability of PC scores was tested as with the IDs before and after mixing the mouse groups such that all individuals were unfamiliar to one another. Only the first principal component remained stable after the mix (one-sided permutation test, n=64 individuals). (d) Scores on the first four PCs were used as predictors of transcriptomic variance in RNA-sequencing data from three different brain regions. This analysis directly mimicked the equivalent analysis performed using the four IDs (PC scores from day 1, 200 shuffled PC score sets; randomization test with n=32 individuals). The top four PCs did not carry more overall transcriptomic information than would be expected by chance.