rmd with the framework to create Mouse brain heatmaps and analyze data generate by Nutil to useable (https://github.com/DaniellaDeWeerd/Nutil-To-Usable).
This code was written using SVGs downloaded from The Allen Brain Atlas http://api.brain-map.org/api/v2/svg_download/100883867?groups=28. Please see their terms of service https://alleninstitute.org/legal/terms-use/ to ensure your use of this code fits both the GNU general pubic license 3 and the Allen Insitutes terms of service. In very brief and overly simplified terms, if this code is used purely for academic purposes, it may be used freely and changed ad libitum, so long as both the image source and this code are properly cited.
The analysis method is robust linear regression on rankit transformed outcomes. This method was chosen as there are often strata with no pathology (true zeros) while another comparative strata has mainly non-zeros. This analysis seeks to answer the question, which regions have more pathology. An alternative method would be zero-inflated beta regression. However, this method is very time-consuming, but potentially more powerful. This app does not implement this method, but it may become available in future versions. Daughter regions are included as fixed-effects as opposed to random to save compute time and because there are only a few daughter regions per parent. A mixed-effects model could yield better estimates of the standard error, but for fast and flexible analyses this was a calculated compromise. For more information on second generation p-values, please view the publications: https://doi.org/10.1080/00031305.2018.1537893 and https://doi.org/10.1371/journal.pone.0188299.
Summary brain heatmaps are either the mean or median. At the parent level this is the mean of means or median of medians.