Hi there!

First, hello to the MIT Tech Review who made me famous for about 15 seconds this month. That’s one of my Facebook comment replies Singhal is pointing at (below) – and hi there Facebook users! It struck me when I saw that how strange the spread of information is, and how much information we read in a heavily digested format in a typical day…

Meanwhile, the LHC is running as normal. No news is good news, for the moment. If you want to hear about some (possibly) exciting physics, I highly recommend the NYT article on a possible hint of new physics from the Tevatron. They do a pretty good job explaining things there, so I won’t confuse you any further. There does seem to be something to it – and now the race is on to find what’s out there at the LHC!! Physicists or the very brave can read the full article here. Notice that their result is “evidence,” * not* “discovery” (or “observation”). That means they’re 99.7% sure the result is new physics – to claim discovery, they’d have to be 99.9999% sure (ok, there needs to be a little more mathematical definition of “sure” to make that precise, but I hope it gives you some idea of what I mean).

*Note:* Fixed my evidence / observation mix up. Thanks for the careful reading…

–Zach

3 sigma does not mean 99.7% the result is new physics! This is a very common and misfortunate misunderstanding. What it really means is that the probability of the null hypothesis fluctuating to produce the observed data is < 100-99.7=0.3%. To decide if it's new physics you have to compare this probability to the probability that new physics actually exists that could produce such a signal. In other words, you need a Bayesian prior for the new physics hypothesis. Put in Sherlock Holmes terms, a statistical fluctuation, however unlikely, may still be the likeliest explanation if all of the new physics alternatives are ruled out on other grounds or are at least even more unlikely.

A true Bayesian! As I said, there needs to be a more mathematical definition of “sure” above – I think something like Bayesian priors are a bit complicated for the average blog reader. But given that we have no real idea what physics alternatives are “likely,” it’s awfully hard to say what the priors ought to be.

I’m not claiming that we’re sure it’s new physics, to be clear. Maybe some blogger braver than I am can have a go at the look elsewhere effect as well…

Thanks for the comment – and I encourage anybody who’s reading this who wants more precision to start with wikipedia articles on Bayesian probability (and a nice TED talk on statistics, available on YouTube) and work from there…