I was noticing the Cooking Dice entry on his blog
http://danielmadison.co.uk/Blog02.html (just go all the way down to the very bottom)
Now, one part caught my attention; it describes the results he made.
As a statistics graduate student, and especially given the fact that I JUST got out of a Nonparametric Procedures midterm, I decided to put that statement to the test. So the setting is this:
Null Hypothesis: p = 1/6 or about 0.16666...
Alternative Hypothesis: p > 1/6
Sample size: 50
Parameter (# of successes): 18
Now the easiest way to take care of this is to find the p-value. In non-stats terms, it's the odds that he'd get 18 or more successes (as in, the face with one pip showing on top). So I went here, since there's no way the tables in my textbook would work:
http://stattrek.com/tables/binomial.aspx
Turns out, P(X >= 18) is equal to 0.00077. That is, the chances that he'd get at least 18 out of 50 rolls such that the desired face (one) shows up is below 0.08%
He should have more confidence in his work; in any case, I'm glad he did the microwave experiments, even though I lost one of my casino dice to trying the same. Oh well, always hope for the future!
If you don't understand, feel free to ask in your reply; I did the stats so you don't have to ;-)
-Sean
http://danielmadison.co.uk/Blog02.html (just go all the way down to the very bottom)
Now, one part caught my attention; it describes the results he made.
Although with this die the chances of landing a 1 are increased, I rolled the die 50 times and only landed 1 - 18 times; which although at 36% is statistically above average (16.6%) I'd be more incline to put that down to the laws of chance than the implementation of a crooked die; to which I see no advantage of taking the risk involved with playing with such an instrument of deception.
As a statistics graduate student, and especially given the fact that I JUST got out of a Nonparametric Procedures midterm, I decided to put that statement to the test. So the setting is this:
Null Hypothesis: p = 1/6 or about 0.16666...
Alternative Hypothesis: p > 1/6
Sample size: 50
Parameter (# of successes): 18
Now the easiest way to take care of this is to find the p-value. In non-stats terms, it's the odds that he'd get 18 or more successes (as in, the face with one pip showing on top). So I went here, since there's no way the tables in my textbook would work:
http://stattrek.com/tables/binomial.aspx
Turns out, P(X >= 18) is equal to 0.00077. That is, the chances that he'd get at least 18 out of 50 rolls such that the desired face (one) shows up is below 0.08%
He should have more confidence in his work; in any case, I'm glad he did the microwave experiments, even though I lost one of my casino dice to trying the same. Oh well, always hope for the future!
If you don't understand, feel free to ask in your reply; I did the stats so you don't have to ;-)
-Sean