Log vs linear debate

New or cumulative? Log vs. linear shouldn’t matter except for scale as long as you know what you are charting.

Well, if the assumption is log-normal distribution then one can regress on log.  Then there are time when log will show the break point because it is not lognormal.  So log is great for proving or disproving lognormal.    Log-normal distributionm look it up. But we would be mostly Poisson if arrivals were normal, and Poisson is closely related to lognormal. In the limit I think Poisson becomes lognormal, though I did not check this.