0

The problem involves two selections one after the other. In the first selection we pick out a set of objects. And then in the second selection we pick out a particular object from this set.

Consider we have 3 sacks each containing $n_1$ red balls (distinct shades), $n_2$ blue balls (distinct shades) and $n_3$ green balls (distinct shades) respectively. Consider $3$ independent random variables $X_1, X_2, X_3$ associated with the number of balls picked from each sack respectively (withput replacement) i.e $X_1$ is the number of red balls picked ($0 \leq X_1 \leq n_1)$. The probabilities of picking a ball associated with each colours are $p_1,p_2,p_3$ respectively.

Now that we have defined how to pick a set of balls which itself is random, we define how to pick a ball of particular color and particular distinct shade form this set or not pick a ball at all.

The probability of picking a ball of particular distinct shade and colour, say a particular shade of RED without loss of generality (present in the set is): $$q_{rl} =\frac{1}{\epsilon + X_1+X_2+X_3}$$

and not picking any ball is:

$$q_n=\frac{\epsilon}{\epsilon + X_1+X_2+X_3}$$

Now what we want to find is the probability of picking up a ball which is RED (shade doesn't matter). Thus the probability for that will be: $$\sum_l q_{rl}$$ i.e the sum of all probabilities of picking any shade of red denoted by $l$ from the constructed set.

What we need is the value of $\mathbb E \big[\sum_l q_{rl} \big]$ i.e the expectation of the probability that the final ball in our hand is red.

It is given that:

$$\mathbb E \big[\sum_l q_{rl}\big] = \mathbb E\bigg[\frac{D_1}{\epsilon + D_1+D_2+D_3}\bigg]$$

where:

$$D_i = Binomial(n_i,p_i)$$

Intuitively this seems clear but I cannot show this formally. I also could not find any resorce to deal with division of random variables except this, which seems insufficient here.

This is intuitive because, one way to write the probability of picking a ball of particular distinct shade and colour from our random set is (say RED):

$$q_{rl} =\frac{C_{rl}}{\epsilon + \sum_{i=1}^{n_1}C_{ri} + \sum_{j=1}^{n_2}C_{bj} + \sum_{k=1}^{n_3}C_{rk}}$$

where $C_{rl}$ will be $Ber(p_1)$, similarly if we had $C_{bl}$ it would be $Ber(p_2)$. Thus what we have is that picking a ball of particular color (say RED) is sum of Bernoulli random variables divided by sum of Bernoulli random variables $\frac{X_1}{\epsilon + X_1+X_2+X_3}$ which intuitively should be related to Binomial random variables, somehow.

DuttaA
  • 205
  • 1
  • 9
  • Your description is rather ambigous. For example, what do you mean by "probability of not picking any ball"? If you put a hand into a bag to pick a ball and take one out, how can you "not pick" a ball at the same time? Probability of picking a ball of a particular color is $X_i / \sum_j X_j$, you have $X_i$ balls of this color and $X_1 + X_2 + X_3$ balls in total. Hard to comment on your calculations as the whole procedure you are describing is not very clear. – Tim May 13 '21 at 07:13
  • @Tim the second part is not about picking from the sack. That was the 1st part. The second part is picking form a set, you can decide to pick up a ball or not pick up anything at all. – DuttaA May 13 '21 at 10:24
  • Then the missing information is about the probability of not picking anything at all, I assume this is given? – Tim May 13 '21 at 10:31
  • @Tim The probability is given of not picking any ball. From my persepective all tge probabilities seem well defined and summing upto 1. What is the missing information? – DuttaA May 13 '21 at 10:33
  • The question currently equivocates over whether $X_i$ are random variables or specific counts, and the $q$ terms sum to one only if $X_1$, $X_2$, and $X_3$ sum to three. – jkpate May 13 '21 at 11:05
  • @jkpate $X_i$ are random variable. In the first line Is the overview of the question of picking from up a random set and then picking an object from the set. It doesn't sum to $3$. I have used $2$ subscripts $r,l$. Summing over $l$ will give me $X_1$. So summing over all $q$'s involve summing over 2 indices. One gives the $X_i$, the other sums to 1. – DuttaA May 13 '21 at 11:14

0 Answers0