Re: question for MIT probability class
So I was a bit bored and decided to solve this problem by hand, but I arrived at a different answer than the above posters. I would appreciate if someone told me where I went awry.
First, consider the possible arrangements of four green balls in four urns. There are four different overarching categories: 3 empty urns, 2 empty urns, 1 empty urn, and 0 empty urns.
If there are 3 empty urns, there are 4 possible configurations:
4000
0400
0040
0004
If there are two open urns, there are 18 possible configurations:
3100
3010
3001
1300
0310
0301
0031
1030
0130
1003
0103
0013
2020
2200
2002
0220
0202
0022
There are 12 possible configurations with 1 open urn:
2110
2101
2011
1201
1210
0211
1120
1021
0121
0112
1012
1102
Finally, there is one configuration with 0 open urns:
1111
This gives us 35 different configurations of green balls among the 4 urns, with 4/35ths leaving 3 open, 18/35ths leaving 2 open, 12/35ths leaving 1 open and 1/35 leaving 0 open urns.
Now, consider the configurations of the 6 black balls among the 4 urns:
Again, we have 4 possible cases, leaving 3, 2, 1 or 0 urns open.
There are 4 configurations leaving 3 open urns:
6000
0600
0060
0006
There are 30 configuations leaving 2 open urns:
5100
5010
5001
0501
0510
1500
1050
0150
0051
0105
1005
0015
4200
4020
4002
2400
0420
0402
2040
0240
0042
2004
0204
0024
3300
3030
3003
0330
0303
0033
There are 40 combinations leaving 1 open urn:
4110
4101
4011
1401
1410
0411
1041
1140
0141
1104
1014
0114
3210
3201
3102
3120
3021
3012
0321
0312
1320
1302
2310
2301
0132
0231
1230
1032
2031
2130
0123
0213
1023
1203
2103
2013
2220
2202
2022
0222
Finally, there are 10 configurations leaving 0 open urns:
3111
1311
1131
1113
2211
2121
2112
1221
1212
1122
This gives us 84 possible configurations of 6 black balls among 4 urns, with the following distribution:
4/84 leave 3 open urns.
30/84 leave 2 open urns.
40/84 leave 1 open urn.
10/84 leave 0 open urns.
Now it is not too difficult to figure out how many configurations have no green and black balls in the same urn.
First, if all the green balls are in one urn (p = 4/35):
4/84 times all the black balls are also in one urn, and 3/4 of the time they will not overlap. So, 4/35 * 4/84 * 3/4 = 12/2940.
30/84 times the black balls will be in 2 urns, leaving two open. 14 out of these 30 configurations will not overlap with any given placement of the green balls. (e.g., if there are four green balls in the first urn, and the black balls are distributed among 2 urns, there is a 14/30 chance that there are no black and green balls in the same urn.) Therefore:
4/35*30/84*14/30 = 56/2940.
Last, when the green balls are in a single urn, 40/84 times the black balls will be distributed among 3 urns, leaving 1 open. 1/4 of those times, the open urn will be the urn in which the green balls are placed. Therefore:
4/35 * 40/84 * 1/4 = 40/2940.
So, P(no overlap|all green balls in a single urn)=P(no overlap|all green balls in a single urn and all black balls in a single urn) + P(no overlap|all green balls in a single urn and all black balls in two urns) + P(no overlap|all green balls in a single urn and all black balls in three urns)
Or: P(no overlap|all green in a single urn) = 12/2940 + 56/2940 + 40/2940
= 108/2940
Next, we consider the cases when the green balls are distributed among 2 of the four urns. There will be only two configurations of black balls that could avoid overlap given 2 urns without green balls, those in which the black balls were distributed among 2 or 1 urns.
If the green balls are distributed in two urns, there are 6 different general configurations possible: 00XX, 0X0X, 0XX0, X00X, X0X0, XX00. If the general configuration of green balls does not overlap with the black balls, the black balls must be arranged in the inverse configuration (e.g., if green are in the 3rd and 4th, 00XX, black must be in the first and second XX00.) So, looking at the 30 configurations of black balls which leave 2 open urns, we see that, for any generic configuration of green balls, there will be 5 of the 30 non-overlapping configurations of black balls.
So: P(no overlap|green balls in 2 urns and black balls in two urns) =
P(green balls in two urns) * P(black balls in two urns) * P(they don't overlap)=
18/35 * 30/84 * 1/6 =
90/2940
Next, consider the possibility that the green balls are distributed among 2 urns, and the black balls are in only 1 urn. In this case, 2/4 configurations of black balls will be non-overlaping, so:
P(no overlap|green balls in two urns and black balls in one) =
P(green in two urns) * P(black in one urn) * 1/2 =
18/35 * 4/84 * 1/2 =
36/ 2940
The total conditional probability P(no overlap|green balls in two urns) =
90/2940 + 36/2940 =
126/2940
Last, consider the case where the green balls are distributed in 3 urns, leaving one open urn. In this case, the black balls must all be in the open urn, which is only one possible configuration. Therefore:
P(no overlap|green balls in 3 urns) =
P(green in 3)* P(black in 1) * 1/4=
12/35 * 4/84 * 1/4 = 12/2940.
So, the total probability of configurations with no black and green balls in the same urn =
P(no overlap|green balls in one urn) + P(no overlap| green balls in two urns) + P(no overlap|green balls in three urns)=
108/2940 + 126/2940 + 12/2940 =
246/2940 =
.08367 or around 8.4% of all configurations have no overlap between the green and black balls.
|