Enjoy this fitted normal distribution of popcorn popping


Originally published at: https://boingboing.net/2018/06/29/enjoy-this-fitted-normal-distr.html


ooh, now roast some coffee and let me manually record your first and second cracks - that’s it, oh yeah



All that shaking and fussing skewed his numbers all over the place - it’s just that, with a normal distribution, it just keeps on looking normal. Better statistics could be had by using one of those air-blower poppers. Do they still make those?


Sure do, we have had great results with this

and this



Popcorn popping should not be described by a normal distribution. Pops/second should follow a Poisson distribution, waiting time to pop should be an exponential.


Popcorn popping should definitely be approximated by a normal distribution. Modeling it as a Poisson distribution requires us to ignore the real trait variability in things like size, shape, thickness of hull, etc., of individual kernels that will have an effect on expected popping time. Not to mention other sources of variability like the distribution of oil and salt in the pan, distance from the heating element, conductivity of the pan, etc. At the intersection of these a normal distribution is probably your best bet.


1 Like

Waiting time and count data like this are what the exponential and Poisson distributions were invented for, and have been standard applied statistics praxis for over 100 years.

1 Like

This is why I come to BBS.

1 Like

Apparently you have never popped popcorn, because a Poisson process it is not. The events are not randomly distributed in time according to some fixed rate - you put the kernels in the oil, wait a while, then all the kernels pop at once in a narrow time interval. The time to pop is a function of variability in the kernels and heating, not pure stochasticity. This should be modeled by a normal distribution, not a Poisson.


I agree that it is not a Poisson process, since the instantaneous occurance rate is not uniform, but if you take a point process even with with nonuniform rate, and measure event counts over a fixed time interval, the should roughly fit a Poisson distribution. That said, I’m not sure what he’s measuring with the reported pops/second.

1 Like
closed #14

This topic was automatically closed after 5 days. New replies are no longer allowed.