Linear mixed-effect modeling of organ of Corti vibratory tuning curves

Optical coherence tomography has become the most popular approach to experimental measures of sound-induced vibrations within the mammalian cochlea. Because it is relatively easy to use and works in the unopened cochlea, the measurement of vibratory tuning curves has become highly reliable, and averaging data from multiple animals in different experimental cohorts is now possible. Here I tested a modern statistical approach to compare cohorts for differences in the magnitude and phase of vibration. A linear mixed-effect approach with first, second, third, and fourth-order models to fit the data was tested. The third-order model best fit both the magnitude and phase data without having terms that did not contribute substantively to improving the R2 or the p-value for the independent variables. It identified a difference between cohorts of mice that were different and no difference between cohorts that should not be different. Thus, this approach provides a way to simply compare a full set of tuning curves between cohorts. While further analyses by the investigator will always be needed to study specific details related to the study hypothesis, this statistical technique provides a simple way for the cochlear physiologist to perform an initial assessment of whether the cohorts are same or different.