Predict M45...

[ewmayer]

Please Don't Flame each other over something this silly.

[philmoore]

Ernst, it wasn't intended as a flame, I really just didn't understand what Phil was referring to. Thanks for the explanation, Drew.

I should teach probability sometime and learn the proper terminology, as my knowledge of it is more based on my experience in physics. However, there is an interesting point about Poisson processes, and that is the long tail that skews the mean to be considerably higher than the median. So if the mean ratio between successive Mersenne prime exponents is roughly 1.48 (assuming none have been missed less than the most recent), the median should be 1.48^(ln 2), or about 1.31. Someone please correct me if I have made a mistake, but it seems to me that we would expect the ratio to be less than this about half the time. Looking at the actual data, we see a median of about 1.41, but this may be somewhat skewed by the small exponents. If we consider only the Mersenne primes discovered by computer, the median is about 1.322, which is pretty close to the median of our guesses so far; I think wpolly's guess of M41991811 is the actual median (ratio of 1.29). So as a group, it looks like our guesses are fairly reasonable, although I detect a faint bias toward the "we're due for a big gap" way of thinking.

[ewmayer]

Originally Posted by philmoore

Ernst, it wasn't intended as a flame, I really just didn't understand what Phil was referring to.

No worries - I was mainly trying to find a cute alternative acronymical mapping for "PDF."

as a group, it looks like our guesses are fairly reasonable, although I detect a faint bias toward the "we're due for a big gap" way of thinking.

Part of this is likely due to the "tens tails in a row .. I'm due for a heads..." style of gambler's fallacy, but perhaps there's also a simple conservatism at work here - guess a big gap, and hope to be pleasantly "disappointed."

[billymac00]

from basic fit, "prediction" puts at 8.6E7 assuming no previous gaps

[PrimeCrazzy]

I predict it will be less than 40,000,000 and found in September 2007

[davieddy]

After consulting Chris Caldwell's page
"Where is the next Mersenne Prime"
It seems there is a 50% chance that it will be less than
43,000,000.
However, I don't think he has taken all the testing
up to 37,000,000 which has been done since the page
was last updated.

David Eddy

[davieddy]

50% chance of it being less than 2^43,000,000
judging by Chris Caldwell's page
"Where is the next Mersenne Prime"

But I don't think it takes into account all the
testing of exponents <37,000,000.

David Eddy

[T.Rex]

Where is the next Mersenne Prime
Within figures 1 and 2, I see "waves" around the theoretical curve. Really clear for fig 2. It really does not look as random values above or below the curve. I guess there is a way to check if the values appear to be random or not. Who can provide the theory and do the computation ?
T.

[davieddy]

Originally Posted by T.Rex

Where is the next Mersenne Prime
Within figures 1 and 2, I see "waves" around the theoretical curve. Really clear for fig 2. It really does not look as random values above or below the curve. I guess there is a way to check if the values appear to be random or not. Who can provide the theory and do the computation ?
T.

Much of the page is devoted to confirming that the occurence
of Mersenne primes is a "Poisson" process, i.e. random.

For example, the discovery of M44 didn't alter the "expected"
primes in each range of the status page at mersenne.org
in the slightest.

The "expected new primes" column on this page provides
the easiest way to keep up to date on the probability
(or lack of it!) of finding a new prime.

If the expected number of primes in a range is x, then the
probability of finding none is e^-x

David

[davieddy]

I do see what you mean about "waves" however.

My guess is that if you chose 44 mumbers at random between
one and a thousand, then plotted them in order and drew the
best fit straight line, you would see a similar effect.

David

[drew]

Originally Posted by T.Rex

Where is the next Mersenne Prime
Within figures 1 and 2, I see "waves" around the theoretical curve. Really clear for fig 2. It really does not look as random values above or below the curve. I guess there is a way to check if the values appear to be random or not. Who can provide the theory and do the computation ?
T.

It's our nature as humans to look for patterns in randomness. You may see waves, but if you tried to derive the wavelength, you'd realize there's no real way to characterize the behavior, assuming you have a large enough sample. A fourier transform of the deviation wouldn't reveal discernable pattern.