Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?

Comment An unknown number of flips (Score 1) 634

What a crock. Only one calls the toss. Have you ever actually done a coin-toss?

If you want to be pedantic, neither of them call the toss-- neither candidate was even present. The county clerk of elections both flips the coin (or designates the person who does so), and calls which result goes to which candidate (or designates the person who does so). If the clerk calls heads for one candidate, they are calling tails for the other.

That's the way coin tosses work: calling heads one way also means calling tails the other.

What I find more puzzling, and probably more worthy of attention, is that the media almost instantly put out the story "Hillary won six out of six coin flips"-- but the best reported actual data is that, of the coin flips that were officially recorded, Sanders won five out of six against Clinton, and one out of one against O’Malley. (e.g., this report).
So, was it just coincidence that the media happened to get 6 out of 6 data points about Hillary winning coin flips against Bernie, but zero out of six data points about Bernie winning coin flips against Hillary? What are the odds of that? Well, the same calculation.

The original six out of six number apparently came from the The Des Moines Register, who based their reporting on what had been posted to Facebook (!). In a later story, they stated that the actual found of coin flips was "unknown".

Comment Re:Two-tailed probability distribution (Score 1) 634

No, you're making false assumption. I never said anything about the probability of heads (or tails) coming up 6 times in a row.

Let me explain this again. If you ever have the question of "should I use one-tailed or two-tailed probability distribution in this statistical analysis," here is a clue: always use two-tailed unless there is a very compelling argument not to.

The question is, what are the probabilities of calling a coin toss correctly 6 times in a row.

And two people called the coin toss: Sanders and Clinton. The odds of one of the two of them calling it correctly 6 times in a row is one in 32, not one in 62.


Comment Random or bias? (Score 1) 634

Hahaha no.

The odds of EITHER Bernie OR Hillary getting 6 for 6 would be 1/32. But we are specifically talking about HILLARY, who is the candidate with all the shenanigans.

What you just did here is to insert your own a-priori personal bias into the statistics.

Your personal bias is fine. Just don't confuse it with statistics.

If Bernie won all the coins, it's not because he knows a guy, it's because 1/64 can happen.

Your statement "result X would show shenanigans if it favors candidate Y, but would be explained by random chance if it favors candidate Z" is a statement of personal bias, not a calculation of statistics.

Anyway, other posts say the 6 for 6 wasn't a the whole story anyway.

Yes, exactly. In fact, the best-documented results are from the counties that did their reporting using the election software (about half the counties reported using the software, and half reported by hand). In these counties, Sanders won six coin flips, and Clinton won one.
So, a more interesting question might be this: given that there were about a dozen coin flips, and Clinton and Sanders won roughly equal numbers, what are the chances that the six that Clinton won would be commented on by the news media, while the six that Sanders won would be ignored? Is this random, or is this bias?

Comment Two-tailed probability distribution (Score 1) 634


People make that statistical error all the time. The question is whether to use one-tailed or two-tailed statistical tests. You are proposing using a one-tailed statistical test, but two-tailed is correct here.

The thing that seems anomalous is getting the same result six ties in a row. Whether that's six heads in a row, or six tails in a row, is irrelevant. What you should be testing for is the probability of getting the same result n times in a row, whether that is heads or tails.

http://www.ats.ucla.edu/stat/m... tells more.

But, wait: what if you are about to say "Well, if it had been Bernie that won six times in a row, I'd accept it as chance, but since it's Hillary, I am suspicious"? Wouldn't that be a reason to use one-tailed probabilities?

No. What you just did there was to insert your own bias. The whole point of statistics is to avoid bias.
A good rule is, if you can't decide if you should use two-tailed probability or one-tailed, always use two-tailed.

--in any case, though, if the Washington Post is accurate, there were over a dozen coin flips, and Sanders also won some of the tosses.

Comment update - there were other tosses which Sanders won (Score 5, Informative) 634

Actually, one in 32 odds. The chance that a coin tossed one time lands with the same face up is 1 in 1. The chance that a coin tossed two times lands with the same face up is 1 in 2, etc.

A little over two standard deviations.

However, as Washington Post notes, "see the update below: there were other tosses which Sanders won."

The update states:
Update: The initial 6-for-6 report, from the Des Moines Register missed a few Sanders coin-toss wins. (There were a lot of coin tosses!) The ratio of Clinton to Sanders wins was closer to 50-50, which is what we'd expect.


Comment Dystopia or Utopia: you decide [Re:Target shooting (Score 1) 555

Well, of course that's also the problem with non-smart guns. The difference is that with non-smart guns, the failures are mostly Type I (gun fires when it's not supposed to), while with smart guns, Type I failures are decreased at the expense of an increase in Type II failures (gun doesn't fire when it's supposed to.)

Not necessarily. A gun programmed to scan its video feed, recognize the face of a particular Geoffrey Landis, and shoot - will be called a very smart gun indeed. Such a gun can easily be imagined to have more type-I failures than a 50 year old reliable gun.

Wow, and I'd thought I was the science fiction writer here.

Comment Re:Land Grab (Score 1) 256

Because global warming is something that happens on a very long time scale, and real estate investment is something that happens on a relatively short time scale.

By the middle of the 2100s, the sea level will, if present trends continue, rise enough to flood Miami. But real estate investors look for profits in ten years or less, and sea levels will rise by no more than millimeters over that time scale.

Comment Why anomaly and not average? (Score 1) 256

FYI, honest question here. Why is NOAA so insistent on only releasing data by anomaly and not by actual temperature?

Because the average temperature isn't actually interesting, and not actually terribly useful.

If you want the global average temperature, just calculate the global average temperature of the baseline year, and add the anomaly relative to that baseline year. As should be obvious, the global average temperature is just a baseline shifting the whole curve up or down, and it's simplest to just subtract it out, unless there's some reason you want that absolute number-- and I can't think of any reason you would want that average number.

Different people calculate the average in different ways. The NASA data has the global temperature average in 2013 at 14.6 degrees Celsius-- that would be a good value to use. (If 2013 isn't your baseline year, subtract the anomaly for 2013 to convert to the baseline year you do use-- it's just a baseline shift; the whole curve shifts by the same amount.)

By only looking at the difference from the baseline, you leave out all of the errors that aren't there if you only are looking at the change in temperature at each location, not the absolute temperature.


Comment Which data set [Re:Here is the adjustment] (Score 2) 256

It's better to just trust the satellite record. (Also, it's fairly annoying how infrequently the error bars are included on those temperature graphs).

There are several satellite data sets, and they have a very large number of corrections required to convert radiance to atmospheric temperature profiles. In general, they give tropospheric temperature, not surface temperature.

Which data set do you like?

Comment Target shooting [Re:Two types of Error] (Score 0) 555

The problem with smart guns is that you get the same failure from each path: Somebody gets shot that wasn't supposed to.

Well, of course that's also the problem with non-smart guns. The difference is that with non-smart guns, the failures are mostly Type I (gun fires when it's not supposed to), while with smart guns, Type I failures are decreased at the expense of an increase in Type II failures (gun doesn't fire when it's supposed to.)

This is what we call an engineering trade-off.

Either someone gets shot by your gun when they weren't supposed to, or you get shot by somebody else's because you couldn't shoot them first.

No, of course not. By far the vast majority of gun use is people shooting at targets, at a range or shooting bottles behind their house or whatever. The effect of a type II error would be the shooter cursing at the gun and then cleaning the contacts or adjusting the hand position or cleaning their hand on a towel or whatever the usual fix is.

Since a Type I failure can very often mean somebody in your family gets shot (often by a toddler), whereas a Type II failure is very rarely fatal (since most shooting is at targets), suppressing type I errors in favor of increasing type II failures is likely to be a good trade-off.

I've never, not once in my life, been shot by somebody because I didn't shoot first. Have you? Really?

Comment Re:So...a year with fewer hurricanes = no warming? (Score 1) 256

Forget basic science and the scientific method. I think the main thing he doesn't understand is sarcasm.

Sarcasm is invisible on the internet. There's so much hyperbole on the net that isn't intended as sarcasm that it's completely impossible to tell when it is.

Slashdot commenters really need to start learning this.

Comment Long term, not short term (Score 0) 256

Don't you understand BASIC SCIENCE?

Extreme weather of any kind is evidence of global warming..

Weather is not climate. Repeat this over and over again: weather is not climate.

One warm winter does not mean "global warming is real." Even a record breaking warm year doesn't mean that global warming is real. Global warming is about long term global averages, not about one particular place, and not about one particular year.

A single record-breaking warm year isn't evidence of global warming... but a string of record warm years is. If it were random, it should be as likely that there are record cold years as record warm years. When you see a string of warm years... that's evidence.

Comment Here is the adjustment (Score 4, Informative) 256

Here is the "adjustment" you're referring to:

The recent correction is the difference between the black line and the red line. The temperature rise between 1959 and 2014 is about 0.9C. The adjustment, in the last two years, is just barely large enough to see, about 0.05C. Over the full period analyzed, the new global analysis changed the observed rate of warming from 0.065C/decade to 0.068C/decade, less than the noise.

Really, I need to point out that analyzing data sets is what science does. But, if you actually look at the data, even if you throw out the new corrections entirely, it doesn't make a difference. The corrections didn't change whether warming exists or not.

That image is from this article: http://arstechnica.com/science...
For reference, here is the paper with the adjustments explained: http://www.sciencemag.org/cont...
(Karl, et al., "Possible artifacts of data biases in the recent global surface warming hiatus," Science Vol. 348 no. 6242, 26 June 2015: pp. 1469-1472
DOI: 10.1126/science.aaa5632)

Slashdot Top Deals

I cannot conceive that anybody will require multiplications at the rate of 40,000 or even 4,000 per hour ... -- F. H. Wales (1936)