Hacker News Reader

Comment by tedunangst | Hacker News Reader

Without knowing the percentage of cases that result in plaintiff wins in other cities, you can't draw any meaningful conclusions from that stat.

View on HN

Patent holders win 78 percent of the time [In Marshall], compared with an average of 59 percent nationwide, according to LegalMetric, a company that tracks patent litigation.

What is the variance? skewness?

How are you wanting to measure those for a Bernoulli variable?

In any given case, the plaintiff either wins or loses. Assuming different cases are independent (which is probably at least a decent assumption), there's no more to say about the distribution than the win probability.

You could of course ask about the variance, skewness, kurtosis, 17th moment, etc., of the distribution of damages awarded or length of trial or total lawyers' fees or something. But for the actual win-or-lose figures, the question doesn't make sense.

Well, it kinda makes sense. Define the outcome to be 0 if the plaintiff loses and 1 if the plaintiff wins. Then with a win probability of 0.78 the variance is 0.17, the skewness is -1.35, the excess kurtosis is -0.17, etc., but there's no more information in any of these numbers than was already provided when we learned that the probability was 0.78.

I don't understand what you don't understand, given that you are yourself taking each county/town (not sure what is the atomic unit of judicial division here) as one RV to which you associate a winning rate. There are surely enough of them to get them in bins of winning rate.

Anyway, my point was to highlight that the mean alone is not very informative.

Ah! You were asking about the distribution of win rates across counties or towns or whatever. I completely misunderstood and thought you were asking for more information about the distribution of wins and losses within Marshall.

My apologies.

I am disturbed to see that the great-grandparent of this comment (the one saying "that doesn't make sense for a Bernoulli variable) has attracted 5 new upvotes since the parent of this comment (the one saying "oops, that was a mistake") was posted.

Everybody: Please stop upvoting my wrong comment. It was wrong. At best it may suggest a lack of clarity in the comment it was replying to, but that doesn't deserve 10 upvotes.

Thank you.

(Not that I mind being upvoted as such, but I do wish I saw more correlation between the quality of my comments and the votes they attract.)

I understood differently, that the parent was asking for the distribution of win rates per town.

Deleted comment

Not sure if your whole reply is sarcastic. See my reply to gjm11 if what you need is an explanation of what I wrote.

Well, he is actually right that the town is essentially funded by patent lawsuits. http://www.nytimes.com/2006/09/24/business/24ward.html?_r=0

Doesn't that make the trial unfair? Isn't "no associated interest" a basic marker of a fair trial?

It's very hard to show actual jury bias.

Though i bet if the anti-patent groups got together and bought marshall a new football stadium, the win rates would go down ....

Well, are the claimants buying them a new football stadium? The Lemley paper I posted below certainly seems to indicate no great preference for plaintiffs over defendants.

Thats a tough one for an appeal though, showing the jury was unfair.

Here's a much more detailed analysis with a lot more data and nuance:

Mark Lemley, "Where to file your patent case"

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1597919

EDT is not actually at the top for claimant win rates.

Reply on news.ycombinator.com