Hierarchical Bayesian modeling of multi-region brain cell count data

We can now collect cell-count data across whole animal brains quantifying recent neuronal activity, gene expression, or anatomical connectivity. This is a powerful approach since it is a multi-region measurement, but because the imaging is done post-mortem, each animal only provides one set of counts. Experiments are expensive and since cells are counted by imaging and aligning a large number of brain sections, they are time-intensive. The resulting datasets tend to be under-sampled with fewer animals than brain regions. As a consequence, these data are a challenge for traditional statistical approaches. We demonstrate that hierarchical Bayesian methods are well suited to these data by presenting a ‘standard’ partially-pooled Bayesian model for multi-region cell-count data and applying it to two example datasets. For both datasets the Bayesian model outperformed standard parallel t-tests. Overall, the Bayesian approach’s ability to capture nested data and its rigorous handling of uncertainty in under-sampled data can substantially improve inference for cell-count data. Significance Statement Cell-count data is important for studying neuronal activation and gene expression relating to the complex processes in the brain. However, the difficulty and expense of data collection means that such datasets often have small sample sizes. Many routine analyses are not well-suited, especially if there is high variability among animals and surprising outliers in the data. Here we describe a multilevel, mixed effects Bayesian model for these data and show that the Bayesian approach improves inferences compared to the usual approach for two different cell-count datasets with different data characteristics.