Hi Joshua,

the boxplot method plots times on a log scale, for good reason. I think that the multiple comparisons should also be done on the log scale, because otherwise the cld column of the printed output is often completely dominated by the outliers (let alone normality assumptions ...), and can be in amazing contradiction to plot results. It would be helpful if there were at least an option for requesting the log scale for the tests, ideally defaulting to the log scale.

In line with this, it would also be a good idea to include the geometric mean in the statistics provided by summary.microbenchmark.

Best, Ulrike

P.S.: I initially posted this in the old repository, because Google pointed me there. Perhaps it would be good to have a pointer from there to here?