However, if the evident empty corner is a nonrandom effect then the statistical significance of the regression gradient could be just an artefact of the absence of points in the empty corner-as opposed to a ‘‘true’’ underlying linear relation through the central portion of the data.