SNR in MASIs over Time

Note.—The linear mixed model was fit by using lme4 (a software package for the R statistical programming language) by means of restricted maximum likelihood, with rat knee as a random effect. We inserted a random effect for each rat knee to account for the within-observation correlation over time. Knees could differ substantially for each rat, so the best modeling strategy was to model the variation between each knee as a random effect. The results are nearly identical when each rat rather than each knee was used as a random effect. The number of observations was 42, the number of groups was 14 (at three time points), the log likelihood was -100.575, the deviance was 201.150, the Akaike information criterion was 217.150, and the Bayesian information criterion was 231.051.

^*Effects relative to baseline factor “viable.”