Biased image cropping and non-independent samples

Abstract

Any figure in a research article will typically represent only a small portion of the total data gained by a researcher for that experiment, and it is therefore key that the figure accurately reflects what was found overall. Furthermore, if individual observations form clusters with differing mean properties, those individual observations would not represent independent samples from the populations being compared. In this example, the question of how to fairly represent and treat image data is addressed.