Errata for Ghahramani.

Remark 1.3.  There is no reason to think that in families with three children, the 8 outcomes are all equally likely,
even approximately, because the decision whether or to have more children is likely to be influenced by the
sexes of the existing children.

Exercise 2.3.6 is poorly worded.  It should be "To be able to directly translate texts from any one of these languages into
any other, ..."

Example 3.5 is ambiguous.  "Given that three of them are spades" has at least three possible interpretations that I can see:
(1) "Given that exactly three of them are spades"
(2) "Given that some particular three of them, chosen before looking at the results (e.g. the first three), are spades"
(3) "Given that at least three of them are spades"
Interpretation (1) makes the question "what is the probability that the remaining five are also spades" trivial,
while under interpretation (3) "the remaining five" doesn't make sense, so I'd say the most reasonable interpretation
would be (2).  However, the solution given is for interpretation (3).


Example 3.19.  If the scientist's testimony is interpreted as P(D | I) = 10^(-9) (and P(D | G) = 1), then it should not be 
"the probability of the sample having come from someone other than the suspect is 10^(-9)" (which the scientist would not 
be able to  know anyway), it should be "the probability of the DNA of any given innocent person matching the sample is 
10^(-9)".   Actually that can't be truthful testimony either: if the  guilty party happens to have an identical twin, 
that identical  twin will certainly match the sample as well.  And even for  non-identical close relatives, I think the 
probability of a match would usually be considerably more than 10^(-9).  Perhaps "the probability
of a randomly chosen person matching the sample is 10^(-9)". 

Example 3.31.  This ignores the obvious, easy method (which works for both with- and without-replacement
versions): there are 16 relevant cards, of which 4 are aces and 12 are face cards.  The first relevant card to
appear is equally likely to be any of the 16.  Therefore the probability that the first relevant card is an ace is
4/16. 

Exercise 3.3.10.  You might note that the fraction of babies that are male is in fact significantly
greater than 1/2 in China these days, indicating that there's more going on than flipping a fair coin.

Exercise 4.3.13 and 4.3.14. "Count the letter Y as a consonant" belongs in 4.3.13, not 4.3.14.


Exercise 7.2.8.  Delete the second "is normal" in the first
sentence. 

Sec. 7.3 Relationship between Exponential and Geometric (p. 289)
and exercise 7.3.14 p. 291.  [X] is not geometric, [X]+1 is.  Note
that [X] can take the value 0, while the geometric (as defined
in this text) only takes values >= 1.

Exercise 11.3.20.  If it's supposed to follow from Chebyshev, the inequality should say P(0 < X < 2n+2) >= n/(n+1), not >.


Exercise 11.5.8.  We need to assume that (1) an employee's guest can't be another employee, and 
(2) the random variables for different employees are independent.  Both are rather doubtful assumptions.