Abstract
If many changes are necessary to improve the quality of neuroscience research, one relatively simple step could have great pay-offs: to promote the adoption of detailed graphical methods, combined with robust inferential statistics. If neuroscience experiments often lead to large, complex and multidimensional data, most analyses rely on t-tests and ANOVAs on means. These analyses are of limited sensitivity to differences among distributions and perform poorly for data that do not conform to the tests' assumptions. Results also rarely show the underlying distributions by using inappropriate graphical representations such as bar graphs. Here we present powerful alternative tools that provide both detailed illustrations of effects and robust inferences. These tools quantify how distributions differ and can be applied at different levels of analysis. We describe this graphical approach to group comparisons, and provide implementations in R and Matlab.
