-
Notifications
You must be signed in to change notification settings - Fork 74
Linear Mixed Effects Modeling
We commonly work on experiments involving repeated measures. Examples include responses measured in cellular assays where the repeated measures may occur over multiple experimental batches, multiple plates, multiple wells within a plate etc. Data derived from behavior of mice assayed over consecutive trials would be another example of experiments involving repeated measures. The distinctions between biological and technical replicates become less clear in these scenarios. Standard statistical tests like t-tests, ANOVA are not applicable. Instead it requires the use of more sophisticated approaches, such as Linear Mixed Effect Modeling.
In this course, we will go over the motivation for the use of these models, the underlying description and assumptions behind them, example scenarios and code using R.
We will go over these slides during the workshop. The code used to fit the various linear mixed effects models discussed in the slides is in this RMarkdown file and the corresponding "knitted" html version is here.
Our core led the development of an R package, RMeDPower2. The functions we developed are based on linear mixed effects models. This is a vignette describing the tool. The package can be installed using the instructions provided on the the README page of the github repo linked to above. RMeDPower2 is an R package that provides complete functionality to analyse data coming from repeated measures experiments, i.e., where one has repeated measures from the same biological/independent units or samples. RMeDPower2 helps test the modeling assumptions one makes, identify outlier observations, outlier units at different levels of the design, estimates statistical power or perform sample size calculations, estimate parameters of interest and also to visualize the association being tested. The functionality is limited to testing associations of one predictor (continuous or categorical, e.g., disease status or brain pathology) along with one another covariate (e.g., gender status) in the context of hierarchical or crossed experimental designs.Please feel free to use this package in your own work and give us feedback, ask questions etc.