When do you use binomial and poisson distribution? How to compute for big factorials?

91 Views Asked by At

Player 1 scored 10 points in a hockey game. Player 2 has the most points of any hockey player in history, averaging 2.629 points/game in the 394 regular season games played in 5 seasons '81-'82 to '85-'86. Computer the probability that Player 2 had at least 10 points in at least one of the 394 games.

Textbook Solution:

Poisson Distribution Formula: $$ {\rm P}(Y \geq 10) = \sum\limits_{y = 10}^\infty {\frac{{e^{ - \mu } \mu^y }}{{y!}}} =1-\sum\limits_{y = 0}^9 {\frac{{e^{ - \mu } \mu^y }}{{y!}}} = 1-0.9996=0.0004. $$ Binomial Distribution Formula: $$ {\rm P}(X \geq 1) = 1-{\rm P}(X=0) = 1-C^{394}_0 (.0004)^0 (1-.0004)^{394} = 1-0.8542=0.1458. $$

The first step is to use poisson distribution formula since the mean is given to find at least 10 points in a game in one of the 394 games played (space).

I don't understand why we need to further use the binomial distribution formula when the probability of one game having at least 10 points is 0.0004?

Also...how do I calculate big factorials for the binomial formula? I seem to get an error. I looked it up and I think Stirling’s formula can be used. This isn't covered in the course content and unsure if there is an easier way to calculate big factorials...

1

There are 1 best solutions below

0
On BEST ANSWER

The question is asking for the probability that Player 2 has at least one game in which he scored at least $10$ points, among the $394$ games played. If you knew the probability that he would score at least $10$ points for any single game, then it's clear that the desired probability would be found through a binomial distribution calculation, just like if I asked you, if you roll a die $394$ times, what is the probability you would obtain at least one six--you'd need to know the probability of obtaining a six for a single roll.

But you don't know the single-game probability of Player 2 scoring at least $10$ goals; instead, you are given the average goals per game, which is $2.629$. That means you have to use another model, one that applies to a single game's performance. To this end, we treat this number as an average rate of goals per game, and the random number of goals is modeled by a Poisson distribution, thus letting us compute the single-game probability of scoring at least $10$ goals.

Regarding your second question, you do not actually need to compute such a large factorial. Since $\Pr[X = 0] = \binom{394}{0} (0.0004)^0 (1 - 0.0004)^{394 - 0}$, we can see that $$\binom{394}{0} = \frac{394!}{0! (394-0)!} = \frac{394!}{394!} = 1.$$ It's only when $n$ is large and $x$ (and $n-x$) are not small that the binomial coefficient $\binom{n}{x}$ is difficult to calculate exactly; e.g., $\binom{394}{2} = \frac{(394)(393)}{2}$ is still easy, but $\binom{394}{157}$ is not. In such a case, Stirling's approximation is appropriate.