# Probability that the ball is red

There is a bag with N balls. Some of them are black and some are red. In the bag there can be 0 red balls or 1 red ball or 2 red balls or N red balls. It is given that each of these N+1 options is equally likely. Now, I pick up P balls from the bag and found that R of them are red. After this I pick up another ball from the bag, what is the probability that this ball is red?

-
Are you familiar with Bayes' Theorem? –  Gerry Myerson Oct 7 '12 at 23:23

Like Alex, I will assume sampling without replacement.

The probability of there having been X red balls out of N initially and R red were found when P were drawn is $$\frac{1}{N+1} \frac{\binom{X}{R} \binom{N - X}{P-R}}{\binom{N}{P}}$$ and, given that, the probability the next ball is red is $\frac{X-R}{N-P}$, which only makes sense (and is between $0$ and $1$) if $R \le X \le R+N-P$ so the result is:

$$\frac{\sum \left(\frac{X-R}{N-P}\right)\frac{1}{N+1} \frac{\binom{X}{R} \binom{N - X}{P-R}}{\binom{N}{P}} }{\sum \frac{1}{N+1} \frac{\binom{X}{R} \binom{N - X}{P-R}}{\binom{N}{P}}}$$

$$= \frac{\sum (X-R) \binom{X}{R} \binom{N - X}{P-R} }{(N-P)\sum \binom{X}{R} \binom{N - X}{P-R} }$$

where the sums are over the possible values of $X$, i.e. $\sum_{X=R}^{R+N-P}$.

-
Thanks, Henry! I now understand the question. Yes, the balls are sampled without replacement. That's why I think the probability of there having been X red balls out of N initially and R were found when P were drawn would not be $$\frac{1}{N+1} {P \choose R}\left(\frac{X}{N}\right)^R\left(\frac{N-X}{N}\right)^{P-R}$$ It should be $$\frac{\binom{X}{R} \binom{N - X}{P-R}}{\binom{N}{P}}\cdot \frac{1}{N+1}$$ as @Alex said. –  Rohit Agarwal Oct 8 '12 at 7:44
@Rohit: Answer edited accordingly –  Henry Oct 8 '12 at 9:00

Try using conditional probability : $\mathrm{Pr}$(($P$+1)th is red) = $\mathrm{Pr}$(($P+1$)th is red | $R$ of $P$ were red)*$\mathrm{Pr}$($R$ of $P$ were red)

-

This answer assumes you sample without replacement.

Denote $X_1$ the number of red balls before the first sampling. $X_1$ follows discrete Uniform distribution with parameter $\frac{1}{N+1}$ Now use Bayes law:

$P($sample $R$ red balls|$X_1$ red balls in the basket)$P$($X_1$ red balls in the basket)

For the conditional probability use Hypergeometric distribution: $$\frac{\binom{X_1}{R} \binom{N - X_1}{P-R}}{\binom{N}{P}}$$ Hence, the probability to sample $R$ red balls in the first sample is $$P(S_1)=\frac{\binom{X_1}{R} \binom{N - X_1}{P-R}}{\binom{N}{P}} \cdot \frac{1}{N+1}$$ After this we have $X_2 =(X_1-R)I_{R \leq X_1}$ red balls left in the basket (since obviously $R$ cannot be larger than $X_1$, hence the indicator function) . Using Bayes law again, the probability to sample one red ball ($S_2$) is
$$P(S_2)=P(S_2|S_1)P(S_1)=\frac{(X_1-R)I_{R \leq X_1}}{N-P-(X_1-R)I_{R \leq X_1}} \cdot \frac{\binom{X_1}{R} \binom{N - X_1}{P-R}}{\binom{N}{P}} \cdot \frac{1}{N+1} \\ =\frac{X_2 I_{X_2 \geq 0}}{N-P-X_2 I_{X_2 \geq 0}} \cdot \frac{\binom{X_1}{R} \binom{N - X_1}{P-R}}{\binom{N}{P}} \cdot \frac{1}{N+1}$$

-