Slashdot Mirror
Judge Rodney Gilstrap Sees A Quarter Of The Nation's Patent Cases (vice.com)
derekmead quotes a report from Motherboard: Since taking the bench in 2011 -- moving literally across the street from his law office into the district courthouse -- Judge Rodney Gilstrap has become one of the most influential patent litigation judges in the country. In 2015, there were 5,819 new patent cases filed in the US; 1,686 of those ended up in front of Judge Gilstrap. That's more than a quarter of all cases in the country; twice as many as the next most active patent judge. This busy patent docket didn't blossom overnight, and it's not some strange coincidence. Due to some unique rules around intellectual property filings, patent holders can often file their lawsuits at any district court in the country, even if neither the plaintiff nor the defendant is based there. By introducing a list of standing court orders and local regulations, the Eastern District of Texas (and, in particular, Gilstrap's division of Marshall) has become the court of choice for many plaintiffs, especially non-practicing entities, often referred to as patent trolls.
I'm really tired of this right-wing meme. Slate did a detailed analysis of this issue, and found that while a disproportionate number of Ninth Circuit rulings were overturned by the Court, an even higher proportion from the Third and Fifth Circuits were overturned. Out of 6,387 on-the-merits decisions in 2006, the Supreme Court reversed 18. That's less than 0.3 percent of their on-the-merits decisions (and, of course, an even smaller proportion of the total Ninth Circuit decisions).
The likelihood of any given Ninth Circuit decision being overturned is vanishingly small.
Wow, that's a shockingly egregious example of how to mis-use statistics.
The "analysis" looked at such a short time period, there were only 64 cases the SCotUS took appeals from a U.S. Court of Appeals. Just by pure chance alone, you'd expect a 12.5% variance in their decisions (that is, whether a case where the SCotUS or USCoA would disagree or agree came up by random chance). The variance is even higher for the individual Federal Courts. The SCotUS only reviewed 24 cases from the 9th USCoA, so the expected random variance is 20%. That is, the 9th Circuit Court's reversal rate was 90.5% +/- 20%. By giving percentages to one decimal place, the article gives the impression that it's doing a precise statistical analysis with 3 significant figures of accuracy, when in truth there's less than one significant figure of precision here.
The analysis doesn't provide raw data for the 3rd and 5th Circuit Court cases so I can't calculate proper variances. But the remaining SCotUS cases (73 - 24 = 49) divided over the remaining 11 Circuit Courts (49 / 11 = 4.45 per Court) gives a best case variance of 47%. If a Circuit Court got fewer than 4.45 cases, its random variance would be even bigger. The uncertainty due to just random chance for this sample size is bigger than the difference from the 9th Circuit's overturn rate, meaning these numbers don't show anything. If the author had turned this article in as his homework for a statistics course, he would've gotten an F.
To demonstrate a trend (or lack of one) with statistics, you need a large enough sample to eliminate random variance. That's why statistical fields like drug approval studies, epidemiology, etc. are so difficult, slow, and expensive. You need a minimum sample on the order of 400-1000 to really start proving a statistical trend is emerging. This article is just some guy punching numbers into his calculator, hitting the % button, and is happy they (randomly) came out the way he wanted so he published them.