Errata

Introduction to Probability, 2nd Edition

Dimitri P. Bertsekas and John N. Tsitsiklis
Athena Scientific, 2008
(last updated 7/27/2021)

• p. 48, line −5: Change “n-element subset” to “n-element set”

• p. 79, line 11: Change “if k << n” to “k = 0, 1, . . . , n”
• p. 95, middle paragraph: Adjust the size of the left brackets
• p. 123, line 11: The solution to this problem was inadvertently included (Problem 2.19 is not a starred
problem). Moreover the solution given in the text for part (a) is incorrect. The correct solution is given
in the on-line solutions.
• p. 161: In Fig. 3.13 change “q”Qto “θ”
n Qn
• p. 192, line 9: Change twice “ i=1 E[Xi2 ]” to “ i=1 (E[Xi ])2 ”
• p. 215: line 4 of the caption of Fig. 4.8, change “=” to “≈”
• p. 226, line 9: Change “E[X]” to “E[X̂]”
• p. 230, line 13: Change “M (s) = E[ssx ]” to “M (s) = E[ssX ]”
• p. 281, line −1: Change “P (Y ≥ ) = P (X1 ≥ , . . . , Xn ≥ ) = (1 − )n ” to “P (Y ≥ ) ≤ P (Yn ≥ ) =
P (X1 ≥ , . . . , Xn ≥ ) = (1 − )n ”
• p. 307, line −4: Change “Gary” to “Garry”
• p. 314, line −13: Change “until” to “up to and including”
• p. 317, line 9: Change “≈ P(A ∪ B)” to√“= P(A ∪ B)”
• p. 445, line −14: “1/5” should read “1/ 5”
• p. 454, line 10: “Θ̂” should read “Θ̂L ”
• p. 480, line −4: Change “θi ” to “θ1 ”
• p. 481, line −4: The formula for θ̂1 should be changed to

n
!
1 1X
θ̂1 = xi yi − ax̄ȳ ,
b − ax̄2 n i=1

where Pn
nσ02 σ 2 + σ12 i=1 x2i
a= , b= .
σ + nσ02
2 nσ12
Note that the non-Bayesian formula onP p. 477 is recovered if we let σ0 and σ1 go to infinity. In that
n
case, a approaches 1, and b approaches i=1 x2i .
• p. 511, line −14, −15: Change three occurrences of e−θ to e−nθ
• p. 513, line 2: Delete ≤ 1

Corrections to the 3rd printing:

The following are corrections to the third printing of the book, and have been fixed in subsequent printings
(Nov. 2013 and later).

• p. 79, line −2: Change “Polish” to “Prussian”

• p. 279, line −4: Change “05675” to “0.5675”
• p. 319, line −5, 6: Change “events at different times are independent, it follows that the random variables
L1 , L2 , . . . are independent.” to “events relating to arrivals at different times are independent, it follows
that the events L1 , L2 , . . . are independent.”
• p. 385, figure caption: “Problem 11” should read “Problem 13”
• p. 427, line −14: Change “= c(k)pΘ (i) pki (1 − pi )n−k ” to “= c(k)pΘ (i) nk pki (1 − pi )n−k ”

1
 • p. 427, lines −1, −2, p. 428 line 1: Change equation to

P(error) = P(Θ = 1, X > k ∗ ) + P(Θ = 2, X ≤ k ∗ )

n   k∗  
X n k n−k
X n k
= pΘ (1) p1 (1 − p1 ) + pΘ (2) p (1 − p2 )n−k
∗
k k 2
k=k +1 k=1
∗   !
n   k
1 X n k n−k
X n
k n−k
= p (1 − p1 ) + p (1 − p2 ) .
2 ∗
k 1 k 2
k=k +1 k=1

• p. 428, line 2: Delete “where c(k) is a positive normalizing constant.”

• p. 496, line −8: “(i)-(ii)” should read “(i)-(iii)”
• p. 499, item (iv): three occurrences of “L(X)” should be replaced by “S”
• p. 519, line −2: “95%” should read “5%”
• p. 520, line 6: “95%” should read “5%”

Corrections to the 1st and 2nd printing:

The following are corrections to the first and second printing of the book, and have been fixed in the third
printing. Books from the 2nd and 3rd printing can be identified by the entry “Second printing” or “Third
printing” below the ISBN number in the copyright page in the front.

• p. 10, line +3: “(A2 ∪ A2 )” should read “(A2 ∪ A3 )”

• p. 74, line −6: “if x is any possible value of X” should read “for any real number x”
• p. 79, first displayed equation: replace “k = 0, 1, . . . , n,” with “if k  n,”
• p. 81, l. 2: “view it” should read “view Z”
• p. 105. In the first two displayed equations, “px|Ai (x | Ai )” should read pX|Ai (x) (three times)
P
• p. 134, line −3: the second sum should read {y|h(y)=v} pY (y)
• p. 161, 3rd line of footnote: “two sets of perpendicular lines” should read “two perpendicular sets of
parallel lines”
• p. 204, caption of Figure 4.1: “4.20” should read “4.2”
• p. 205, second line of caption of Figure 4.2: “Y ” should read “aX”
• p. 226, line 12: “X = X̃ + X̂” should read “X = X̂ − X̃”
• p. 260, line 9: “SECTION 4.4” should read “SECTION 4.5”
• p. 271, line −1: “level” should read “levels”
• p. 284, line 2: “Some Useful Inequalities” should read “Markov and Chebyshev Inequalities”
• p. 322, line −11: “U is the time of the first arrival” should read “U is the time of the last arrival”
• p. 326, line 2: “SECTION 7.1” should read “SECTION 6.1”
• p. 329, line −12: “SECTION 7.2” should read “SECTION 6.2”
• p. 334, line 16: in the definition of the event A, “fY (y)δ” should read “λδ”
• p. 344, Figure 7.4: remove the superfluous label “r” above the bottom arc
• p. 346, Fig. 7.5: the upper-rightmost transition probability is given as pij , but it should be p1j .
• p. 376, lines 13 and 14: “frequency” should read “expected frequency” (twice)
• p. 378, line 16: “statistics” should read “behavior”
• p. 390, title of Problem 21: “Expected long-term” should read “Long-term expected”
• p. 394, lines 1-3: “expected long-term” should read “long-term expected” (three times)
• p. 396, part (b) of Problem 30: rephrase as “Do the n-step transition probabilities starting from state
1, r1j (n), converge? If so to what values?”
• p. 396, part (c) of Problem 30: rephrase as “Do the n-step transition probabilities starting from state
6, r6j (n), converge? If so to what values?”

2
 1

P X
(1 k)
0.8
P X
(2 k)

0.6
Posterior Posterior
0.4 P(Θ = 1 | X = k) P(Θ = 2 | X = k)

0.2

Number of heads k
0
0 10 20 30 40 50
k

Choose Θ = 1 Choose Θ = 2

Probability of Error
0.5

0.48

0.46

0.44

0.42

0.4

0.38
N
"
0.36 Nmax pi xi
!N"
!N max i=1 pi xi
wi xi ≤W
i=1
0.34 wi xi ≤W xi =0 or 1
i=1 i=1
xi =0 or 1
1
a0.32
b A fX|{X∈A} (x) fX (x) x y µx + σx µy + σy (µx , µy )
0 10 20 30 1 40 3 50
Set E of pairs Efficient Frontier MAPMAP k
Rule Selection3 Threshold
Threshold k̄ k∗
MAP Rule Subgradients
Threshold k̄ Template
γ α βSubgradients
g(x) = H0 or g(x) = H1 fx (d) g
Template
• p. 427, 428: The fonts for figures fx (d) are shown below
Space8.5 of and 8.6 were corrupted. The corrected

2Possible Observations x Calculation Process
figures
Posterior pΘ|X (· |
k2 k
fx (d)
X=
• p. 434, last line: n should x) P(Θ = 1 | X = k) P(Θ = 2Posterior
read | X = k) pΘ|X (· | X = x) P(Θ = 1 | X = k) P(Θ = 2 | X = k)
n
Posterior pΘ|X (· | X = x) P(Θ = 1 | X = k) P(Θ = 2 | X = k)
• p. 474, line −10: “and variance 0.245” should read “and variance Ŝn2 /n” .
p (·; θ ) p (·; θ) p (·; θ ) p (x; θ ) p (x; θ ) ..
m m
• p. 495, line 10: “right-hand side” should
X 1
readX X
“left-hand X
side” 1 X
Choose Θ = 2 Θ
∗ = 1
• p. 497, line −7: for greater generality, “hypothesis “H0 : θ = θ ” ”” should read “hypothesis H0 ”
Hypothesis selection Θ = 2 Θ = 1
Choose Θ = 2 Θ = 1
• p. 508, Iine 12: Item (a) should read: “Let θ = (p1 , µ1 , σ1 , . . . , pm , µm , σm ). Write the likelihood and
fS (s; H0 ) pX (·; θ) pX (·; θm ) pX (x; θ1 ) pX (x; θm )
log-likelihood functions for θ.”
Number of heads k Θ, X Observation X = x Prior pΘ or fΘ
Decision Rule Confidence
Numberinterval
of headsetc
k Θ, X Observation X = x Prior pΘ or fΘ
v log ξ + 1/2
Conditional pX|Θ or fX|Θ Point Estimates Error Analysis etc
MAP Estimate MLConditional
Estimate p
X|Θ or fX|Θ Point Estimates Error Analysis etc

pX (· ; H1 ) pX (· ; H0 ) pX (·; θm ) pX (x; θ1 ) pX (x; θm )

H0 or H1 Number of heads kPosterior pΘ|X or fΘ|X
Θ, X Observation X = x Prior pΘ or
fΘ Posterior pΘ|X or fΘ|X
Direction of Increasing False Rejection Probability False Acceptance
Direction of Increasing
Probability False
sensors Rejection
object Probability
at unknown False
location Θ Acceptance
Probability sensors∂! fobject
(x) at unknown location Θ
∂! f (x) α × y x yi − θ̂0 − θ̂1 xi
Conditional pX|Θ or fX|Θ Point Estimates α ×y Error
y x yAnalysis etc
i − θ̂ 0 − θ̂ 1 xi
3 −x
Slope ĝk = k c k+1
y −x k
Slope ĝk = kθ̂ c k+1
k
1
$1k = δk − #xk+1 − yk #2
Posterior pΘ|X or Posterior fΘ|X (θ | $x) = pΘ|X
δ (θ
− | x)#x −2cy k #2
k k k+1 k
Direction of Increasing False Rejection Probability 2ck False Acceptance
Residual Slope
Probability sensors object at unknown = 0Θ
location