Disclaimer. Though I try to be precise and correct, errors are inevitable.

If you spot an
error, please mail me. Otherwise, use with caution.
1 Inequalities
1.1 Markov / Chebychev
Let X be a non-negative random variable. For > 0,
P[X ]
E(X)

Applying this to [X E[X][ one has Chebychevs inequality:

Let X be any random variable. For > 0,
P[[X E(X)[ ]
Var(X)

2
[7]
1.2 Paley-Zygmund
Let X be a non-negative random variable. Then for any 0 < 1,
P[X > E[X]] (1 )
2
(E[X])
2
E[X
2
]
.
Proof. Using the Cauchy-Schwartz inequality (1.7 below),
E[X] = E
_
X1
{XE[X]}

+E
_
X1
{X>E[X]}

E[X] +
_
E[X
2
] P[X > E[X]].
.
1.3 Azuma (Cherno / Hoeding)
Let (M
n
)
n0
be a martingale with M
0
= 0. Assume [M
n
M
n1
[ 1 a.s. for all n. Then
for > 0,
P[M
n
> ] < exp
_

2
2n
_
,
P[[M
n
[ > ] < 2 exp
_

2
2n
_
.
[1]
1
1.4 Bernstein
Let X
1
, X
2
, . . . , be i.i.d. random variables such that E[X
i
] = 0 and [X
i
[ 1 a.s. Let

2
= E[X
2
i
], and let S
n
=

n
i=1
X
i
. Then for any > 0
P[S
n
> ] < exp
_

2
2
2
n +
2
3

_
,
P[[S
n
[ > ] < 2 exp
_

2
2
2
n +
2
3

_
.
[3]
1.5 Extension of Hoeding
Let X
1
, . . . , X
n
be real-valued random variables, such that
1. For all i, X
i
is not independent of at most d other variables; i.e.,
max
_
[A[

A [n] , X
i
is not independent of X
j
: j A
_
d.
2. For all i, [X
i
[ b.
Let S
n
=

n
1
X
i
. Then for > 0,
P[[S
n
E[S
n
][ ] 2 exp
_

2
2nb
2
(d + 1)
_
.
[13]
1.6 Jensen
Let X be a real valued random variable. If f : R R is a convex function, then
f(E[X]) E[f(X)].
1.7 Cauchy-Schwartz and H older
Let p, q [1, ] be such that
1
p
+
1
q
= 1. Then,
_
[fg[ d
__
[f[
p
d
_1
p

__
[g[
q
d
_1
q
.
The Cauchy-Schwartz inequality is Holder with p = q = 2.
2
1.8 Doobs Maximal L
p
-inequality
Let 0 = M
0
, M
1
, . . . , M
n
be a martingale (or positive sub-martingale). Let M

= max
0kn
[M
k
[.
Then, for p 1 and any > 0,
P[M

]
|M
n
|
p
p

p
,
and for any p > 1,
|M
n
|
p
|M

|
p

p
p 1
|M
n
|
p
.
[14]
1.9 Harmonic-Geometric-Arithmetic Means
Let a
1
, a
2
, . . . , a
n
be positive real numbers. Then,
n

n
j=1
a
1
j

j=1
a
1/n
j

1
n
n

j=1
a
j
.
[10]
1.10 Prekopa-Leindler
Let f, g, h : R
n
R be non-negative integrable functions and let 0 < < 1. If h((1 )x +
y) f
1
(x)g

(y) for all x, y R

n
, then
_
R
n
h(x)dx
__
R
n
f(x)dx
_
1
__
R
n
g(x)dx
_

.
[8]
1.11 Young and inverse Young
Let p, q, r > 0 be such that
1
p
+
1
q
= 1 +
1
r
. Let f L
p
(R
n
) and g L
q
(R
n
) be non-negative
functions. Then,
if p, q, r 1 then |f g|
r
C
n
|f|
p
|g|
q
,
if p, q, r 1 then |f g|
r
C
n
|f|
p
|g|
q
,
3
where C =
CpCq
Cr
, and
C
s
=

_
[s[
1/s
[s

[
1/s

,
for
1
s
+
1
s

= 1.
[8]
1.12 Brunn-Minkowski
Let X, Y be non-empty bounded measurable sets in R
n
. Let s, t > 0 and assume that
sX +tY
_
sx +ty

x X, y Y
_
is also measurable. Then,
Vol(sX +tY )
1/n
sVol(X)
1/n
+tVol(Y )
1/n
.
Equivalently, for any 0 < < 1,
Vol(X + (1 )Y ) min Vol(X), Vol(Y ) .
1.13 Newton
Let a
1
, . . . , a
n
be non-zero real numbers (positive or negative). For 0 j n let
p
j
=
1
_
n
j
_

S{1,...,n}
|S|=j

iS
a
i
.
Then, for all 1 j n 1,
p
j1
p
j+1
p
2
j
.
Corollary. Write
n

i=1
(x +a
i
) =
n

j=0
_
n
j
_
p
j
x
j
.
Then, p
j1
p
j+1
p
2
j
and if the a
i
s are all positive then
p
1
p
1/2
2
p
1/n
n
.
[10]
4
1.14 Four Functions Theorem and FKG
Let L be a nite distributive lattice; i.e. L is a partially ordered set, such that for all
x, y, z L,
There exists a unique element of L that is the maximal lower bound of x, y, denoted
by x y (and called the meet of x, y).
There exists a unique element of L that is the minimal upper bound of x, y, denoted
by x y (and called the join of x, y).
(Distributivity)
x (y z) = (x y) (x z).
For X, Y L dene:
X Y =
_
x y

x X, y Y
_
X Y =
_
x y

x X, y Y
_
.
For a real valued function : L R and X L dene
(X) =

xX
(x).
Four Functions Theorem (FFT). If , , , : L R
+
are four non-negative real valued
functions of L such that for any x, y L
(x)(y) (x y)(x y),
then for any X, Y L,
(X)(Y ) (X Y )(X Y ).
FKG. Let be a probability measure on L such that
(x)(y) (x y)(x y).
Let f, g, h : L R
+
be non-negative real valued function of L such that f, g are increasing
and h is decreasing. Then,
E[fg] E[f] E[g], and E[fh] E[f] E[h].
In other words, f, g are positively correlated, and f, h are negatively correlated.
[1]
5
1.15 Shearer
For a random vector X = (X
1
, . . . , X
m
) and A 1, 2, . . . , m, set X
A
= (X
i
: i A). H()
is the entropy function.
Let X = (X
1
, . . . , X
m
) be a random vector, and let /be a collection of subsets of 1, 2, . . . , m,
possibly with repeats, such that every element of 1, 2, . . . , m is contained at least t sets in
/. Then for any partial order on 1, 2, . . . , m,
H(X)
1
t

AA
H
_
X
A

(X
i
: i A)
_
.
[5]
1.16 Berry-Esseen
Let X
1
, . . . , X
n
be i.i.d. with E[X
i
] = 0. Assume that E[X
2
i
] < and E[[X
i
[
3
] < .
Let N be a normal random variable with mean 0 and variance 1. Set X =

n
i=1
X
i
. Then,

P
_
X >

P[N > ]

P
_
X

P[N ]

n
.
[7]
1.17 Chen-Stein Method (simplied)
Let X
1
, . . . , X
N
be N Bernoulli random variable with p
i
= E[X
i
] > 0, and let S
N
=

N
i=1
X
i
.
Let G = (V, E) be a graph on the vertex set V = 1, . . . , N, such that i, i E is an edge
(self loop) for all i V . Assume that for all i V ,
E
_
X
i

ji
X
j

= p
i
.
(e.g. this holds if X
i
is independent of the set X
j
: j , i.)
Let Z be a Poisson random variable with mean
E[Z] = = E[S
N
] =
N

i=1
p
i
.
Dene
B
1
=
N

i=1

ji
E[X
i
] E[X
j
] ,
6
B
2
=
N

i=1

ji
j=i
E[X
i
X
j
] .
Then,
sup
AN
[P[S
N
A] P[Z A][
1 e

(B
1
+B
2
).
[2]
1.18 Suens Inequality
For a graph G = (V, E) write i j if i, j is an edge. For subsets A, B V , write A B
if there exists an edge between A and B. Thus, A , B means that there is no edge between
A and B. i A means there is an edge between i and some element of A.
Let X
1
, . . . , X
N
be N Bernoulli random variables, and let S
N
=

N
i=1
X
i
.
Let G be a dependency graph of X
i

N
i=1
; i.e. G = (V, E) is a graph on the vertex set
V = 1, . . . , N such that for any two disjoint subsets A, B V , if A , B then the two
families X
i

iA
and X
i

iB
are independent of each other. (In some texts G is called a
superdependency digraph.)
Dene
=
1
2
N

i=1

ji
E[X
i
X
j
]

k{i,j}
(1 E[X
k
])
1
.

=
1
2
N

i=1

ji
E[X
i
] E[X
j
]

k{i,j}
(1 E[X
k
])
1
.
Then,
P[S
N
= 0] e

i=1
(1 E[X
i
]) ,
P[S
N
= 0]
_
1

_
N

i=1
(1 E[X
i
]) .
Note that if X
i

N
i=1
are all independent, then equality holds in both inequalities (choosing
the empty graph).
[12, 16]
7
2 Convergence Theorems
2.1 The Law of the Iterated Logarithm
Let X
1
, X
2
, . . . be i.i.d. with E[X
i
] = 0 and 0 <
2
E[X
2
i
] < . Set S
n
=

n
1
X
i
. Then,
P
_
limsup
S
n
_
2
2
nlog log n
= 1
_
= 1.
[15]
2.2 Monotone Convergence
Let (, T, ) be a measure space. Let 0 f
1
f
2
be a monotone sequence of
non-negative measurable functions. Then,
f = lim
n
f
n
exists a.e., and is a measurable function. Further,
_

fd = lim
n
_

f
n
d.
[6]
2.3 Fatous Lemma
Let (, T, ) be a measure space. Let f
n
be a sequence of non-negative measurable func-
tions. Then,
_

liminf
n
f
n
d liminf
n
_

f
n
d.
[6]
2.4 Dominated Convergence
Let (, T, ) be a measure space. Let f
n
, f, g be measurable functions such that:
(1) f = lim
n
f
n
a.e.,
(2) n [f
n
[ g a.e.,
(3)
_

gd < .
8
Then,
lim
n
_

f
n
d =
_

fd.
Remark: The condition of a.e. convergence in (1), can be replaced by convergence in
measure; i.e. it suces to require that
(1) > 0 lim
n
[f
n
f[ > = 0.
[6]
3 Formulas
3.1 Stirlings Formula
n!

2n
(n+1/2)
e
n

2n
_
n
e
_
n
e
1
12n+1
< n! <

2n
_
n
e
_
n
e
1
12n
For 0 < < 1,
_
n
n
_
=
_
1 O
_
n
1
__
C
1

n
2
nH()
,
where H() = log (1 ) log(1 ), and the constant is C =
1

2(1)
.
Less accurate, but easier to use:
_
n
k
_
k

_
n
k
_

_
ne
k
_
k
3.2
_
1
1
x
_
and e:
For all x 1,
_
1
1
x
_
x
e
1

_
1
1
x + 1
_
x
.
9
3.3 Finite sums of powers
n

k=1
k =
n(n + 1)
2
n

k=1
k
2
=
n(n + 1)(2n + 1)
6
n

k=1
k
3
=
_
n(n + 1)
2
_
2
n

k=1
k
4
=
n(n + 1)(2n + 1)(3n
2
+ 3n 1)
30
[9]
3.4 Permutation xed points
Let D(n, m) denote the number of permutations of 1, . . . , n that have exactly m xed
points. Then,
D(n, m) =
n!
m!
nm

k=0
(1)
k
k!
.
The proof uses the inclusion-exclusion principle, see e.g.
Wikipedia: Random permutation statistics .
4 Complex Analysis
Denote the unit disc in the complex plane by U;
U = z C : [z[ < 1 .
4.1 Schwartz Lemma
Let f : U C be a bounded analytic function with f(0) = 0. Then,
[f(z)[ [z[ |f|

and [f

(0)[ |f|

.
If there exists z ,= 0 such that equality holds for one of the above, then there exists [0, 2)
such that f(z) = e
i
|f|

z for all z.
[4]
10
4.2 Bieberbach Conjecture (De Branges Theorem)
Let f : U C be a conformal map of the unit disc (i.e. f is injective and analytic in the
unit disc). Then for all n 1,
[f
(n)
(0)[ n n! [f

(0)[.
[4]
4.3 Koebe
1
4
Theorem
Let f : U C be a conformal map of the unit disc (i.e. f is injective and analytic in the
unit disc). Then,
_
z C : [z f(0)[ <
|f

(0)|
4
_
=
1
4
[f

(0)[U +f(0) f(U).

Furthermore, let d
f
(z) be the distance of f(z) from f(U). Then,
1
4
[f

(z)[ (1 [z[
2
) d
f
(z) [f

(z)[ (1 [z[
2
).
[4]
4.4 Koebe Distortion Theorem
Let f : U C be a conformal map of the unit disc (i.e. f is injective and analytic in the
unit disc). Then,
[z[
(1 +[z[)
2
[f

(0)[ [f(z) f(0)[

[z[
(1 [z[)
2
[f

(0)[,
1 [z[
(1 +[z[)
3
[f

(0)[ [f

(z)[
1 +[z[
(1 [z[)
3
[f

(0)[.
[4]
References
[1] N. Alon, J. H. Spencer, The Probabilistic Method (2000), John Wiley & Sons, Inc.
11
[2] R. Arratia, L. Goldstein, L. Gordon, Two moments suce for Poisson approxi-
mations: The Chen-Stein method. Ann. Probab. 17 (1989), 925.
[3] N. Berestycki, R. Nickl, Concentration of Measure (2009), Section 1.5, lecture
notes. available here
[4] J. B. Conway, Functions of One Complex Variable II (1995), Springer-Verlag.
[5] T. M. Cover, J. A. Thomas, Elements of Information Theory (1991), John Wiley
& Sons, Inc.
[6] J. L. Doob, Measure Theory (1994), Springer.
[7] W. Feller, Introduction to Probability Theorey and its Applications (1966), John
Wiley & Sons, Inc.
[8] R. J. Gardner, The Brunn-Minkowski Inequality, Bulletin of the American Math-
ematical Society 39 (2002), no. 3, 355405. available here
[9] I. S. Gradshteyn, I. M. Ryzhik, Table of Integrals, Series, and Products (1965),
New York : Academic Press.
[10] G. H. Hardy, J. E. Littlewood, G. Polya, Inequalities (1952), Cambridge University
Press.
[11] F. den Hollander, Large Deviations (2000), AMS.
[12] S. Janson, New versions of Suens correlation inequality. Random Structures and
Algorithms 13 (1998), 467483.
[13] S. Janson, Large deviations for sums of partly dependent random variables. Ran-
dom Structures and Algorithms 24 (2004), no. 3, 234248. available here
[14] D. Revuz, M. Yor, Continuous martingales and Brownian motion (1991),
Springer-Verlag.
[15] A. N. Shiryaev, Probability, Translated by R. P. Boas (1996), Springer.
[16] W. C. S. Suen, A correlation inequality and a Poisson limit theorem for nonover-
lapping balanced subgraphs of a random graph. Random Structures and Algo-
rithms, 1 (1990), 231242.
12