« Convenience charting |
| Error spotting »
A puzzle (from Laurie Snell, via the Gelman blog): are the tiles on this wall random or not? How does one prove something's random? Or that something's not random?
On this blog, we have addressed this question before when we discussed lottery numbers, and suicides.
Reference: Gelman's blog
Posted on Jan 28, 2007 at 06:59 PM | Permalink
That's a nice question! You can't really answer if it was random after the wall is already tiled when you don't know the selection process.
Statistics: Maybe someone of the other commenters has does a significance test but eventually that's no proof because in 5% of cases (using 95%-confidence interval) it could be random even if the test says it isn't and the other way round.
Visual inspection: Maybe you see patterns emerge (there is a T! And there is a flower!) and conclude: This cannot be random! There is a deeper meaning hidden...
Conclusion: It is random if the tiler chose the tiles in random selection. In random selection every tile should have an equal chance to be the next one tiled to the wall.
Jan 28, 2007 at 07:50 PM
I just read the solution posted by Mike A. (Mike Anderson?) on the Gelman blog :)
So thing about it BEFORE reading the comments on the Gelman blog.
Jan 28, 2007 at 07:58 PM
It wasn't me! I'm afraid statistical tests can only determine whether the pattern is NOT random, according to some criterion. Unfortunately, there's always someone faster, smarter, and with a better criterion just around the bend.
Mike Anderson |
Jan 28, 2007 at 10:56 PM
For me at least, it's easier to see the repeating pattern when I rotate the picture 90 degrees clockwise. Then I see clearly that there are three "cycles" of the pattern.
John S. |
Jan 29, 2007 at 05:47 AM
JS: Thanks for the tip! Once I saw it, I was able to also see it "right-side up." The three regularly-spaced "T"s down the right side are easy to see. Every group of tiles has three instances from top to bottom.
Xris (Flatbush Gardener) |
Jan 29, 2007 at 06:50 AM
I still see tetris and I bet if you collapse all the dark green squares down they form a solid block
Steven Citron-Pousty |
Jan 30, 2007 at 01:39 PM
Look closer, tilt the image 90 degrees to the left and cross-eye yourself: the pattern is really a stereogram showing Einstein's profile! I don't think this has ever been done before, a wall-sized, wall-mounted, tile-made stereogram... Truly amazing!
Jan 30, 2007 at 04:51 PM
Sorry Pedro, I don't see it. I can easily get the image to 'pop', but that only proves it is repetative. No Einstein that I see.
Jan 31, 2007 at 02:26 PM
We were once required to verify whether a set of numbers was random as they affected the outcome of a multimilllion dollar experiment. One statistician quit over the idea that it should even be attempted. The rest of us, preferring an income, came up with some really weak tests, which may be as good as it gets. There is even the problem that many sets of random numbers, including sevaral famous tables, are "log evenly distributed" which apparently happens alot in natural random processes.
Bruce weaver |
Feb 02, 2007 at 06:23 PM
I guessed that because none of the pink tiles had an adjacent pink tile, that this was not random. Is that an accepatble approach?
Feb 07, 2007 at 03:36 PM
Absolutely, Sam. That's one of the tell-tale signs. As is evident here, proving something is random is much harder than finding a counterexample of something not random!
Feb 07, 2007 at 09:26 PM
I agree with what you are saying and this is a great article.
Dec 14, 2009 at 08:54 PM
This is only a preview. Your comment has not yet been posted.
The letters and numbers you entered did not match the image. Please try again.
As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.
Having trouble reading this image? View an alternate.
(You can use HTML tags like <b> <i> and <ul> to style your text.)
(Name is required. Email address will not be displayed with the comment.)
Name is required to post a comment
Please enter a valid email address