Formula Sheet

1. Basic sums:
n i=1 1 i = 2 n(n + 1), n 2 i=1 i

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

2. Binomial coecients: 3. 4. 5. 6. 7. 8.

n n n! k := (nk)!k! = nk , n, k n Sums of them: i=0 k+i1 = n+k , k > n 0, i k n n i ni Binomial theorem: = (a + b)n , n 0, a, b R. i=0 i a b n n 1an+1 i if a = 1, and i=0 ai = n + 1 if a = 1. Sum of a geometric progression: i=0 a = 1a Derivative: f (a) := limh0 f (a+h)f (a) = limxa f (x)f (a) . h xa df dg dg df df dg d d d Dierentiation: dx (f + g) = dx + dx , dx (f g) = f dx + g dx , dx (f g) = dg dx (f g)(x) := f (g(x)) . b b Integration by parts: if g(x) = G (x), then a f (x)g(x) dx = f (b)G(b)f (a)G(a) a f (x)G(x) dx.

n 1 2 3 2 i=1 i = 4 n (n + 1) , n 1. n 0, n+1 = n + k1 , n, k 1. k k n l ml m i=0 i ni = n , m 0, n, l = 0, . . . , m.

9. Taylors theorem: if f, f , . . . , f (n+1) are dened on [a, x], then (using the Lagrange remainder) f (x) = f (a) + f (a)(x a) + +
f (n) (a) n! (x

a)n +

f (n+1) (t) (n+1)! (x

a)n+1 for some t (a, x).

10. Triangle inequalities: ||x| |y|| |x + y| |x| + |y|. x 11. Logarithm: ln(x) = 1 1 dt, x > 0, ln(xy) = ln(x) + ln(y), ln(y ) = ln(y), loga (x) = ln(x)/ ln(a). t 12. Exponential: exp := ln1 , e := exp(1), ex := exp(x), ex+y = ex ey , ax := ex ln(a) , ax+y = ax ay , a > 0.
d ax dx e

= aeax , a R,
i=0

d bx dx a

= b ln(a)abx , a > 0,
i=1

13. Geometric series: 14. Binomial series:

ai =

1 1a ,

iai1 =
1 1+a n

1 1a

ai i=0 i! 2

= ea , limn 1 +
ai i=1 i

a n n 1 1a

= ea , a R. , 1 a < 1.

, |a| < 1,

= ln

n+i1 i=0 i

(1)i ai =
sin x cos x ,

, n 1, |a| < 1.

15. Trigonometric functions: tan x = sin2 x + cos2 x = 1, tan(x + y) = 16. 17.

sin(2x) = 2 sin x cos x, cos(2x) = 1 2 sin2 x = 2 cos2 x 1,

sin(x + ) = cos x, sin(x + 2k) = sin x, cos(x + 2k) = cos x, tan(x + k) = tan x, 2 sin x = cos x, cos x = sin x, +
x5 5! d dx

tan x = 1 + tan2 x,
x2 2!

sin(x + y) = sin x cos y + cos x sin y, cos(x + y) = cos x cos y sin x sin y,
tan x+tan y 1tan x tan y ,

sin x = x

x3 3!

x7 7!

+ , cos x = 1

x4 4!

x6 6!

+ . ,

Gamma function: () = 0 x1 ex dx, > 0, ( + 1) = (), 1 Beta function: (a, b) = 0 xa1 (1 x)b1 dx = (a)(b) , a, b > 0. (a+b)

1 (n + 1) = n!, ( 2 ) =

18. Sterlings formula: n!

2nn+1/2 en as n , n! =

2nn+1/2 e(n+ 12n ) for some t (0, 1).

19. Probability. is the sample space, A, B, are events, P () is probability measure, X, Y, are random variables (rvs), S is the range of X if X is a discrete rv, FX denotes distribution function, fX denotes probability density function (pdf) when X is a continuous rv, and, E() is expectation. (a) P (AB) = P (A)+P (B)P (AB), P (A|B) := P (AB)/P (B), P (B|A) = P (A|B)P (B)/P (A). (b) Total probability: P (A) = (d) E(X) = (e) E(g(X)) =
i

P (A Bi ) =

P (A|Bi )P (Bi ), where {Bi } is a partition of .

(c) FX (x) := Pr(X x). If X is a continuous rv, FX (x) =

fX (u) du.

x Pr(X = x) (discrete), E(X) =

xfX (x) dx (continuous).

xS

g(x) Pr(X = x) (discrete), E(g(X)) =

g(x)fX (x) dx (continuous).

(f) Var(X) := E((X E(X)) ) = E(X )(E(X)) , Var(X +Y ) = Var(X)+Var(Y )+2Cov(X, Y ). (g) Cov(X, Y ) := E((X E(X))(Y E(Y ))) = E(XY ) E(X)E(Y ). (h) Conditional expectation: E(E(Y |X)) = E(Y ), E(Y g(Z)|Z) = g(Z)E(Y |Z). (i) Probability generating function (pgf): if X is a non-negative discrete rv, GX (z) = E(z X ). (j) Moment generating function (mgf): if E(|X|k ) < for all k, MX (t) = E(etX ). (k) Laplace-Steiltjes transform (LST): if X is a non-negative rv, LX (t) = E(etX ), t 0. (l) Characteristic function (cf): if X is any rv, X (t) = E(eitX ), t R (here i = 1). 1

20. Discrete distributions: Here X is a discrete rv taking values in a denumerable set. The mean, variance and probability function are listed, together with the pgf G(z) = E(z X ), |z| 1. Constant Pr(X = c) = 1, E(X) = c, Var(X) = 0, G(z) = z c . Binomial (B(n, p): 0 < p < 1, n 1) E(X) = np, Var(X) = np(1 p), Pr(X = j) =
n j

pj (1 p)nj , j {0, 1, . . . , n},

G(z) = (1 p + pz)n .

The Bernoulli distribution is the special case B(1, p). Poisson (Poisson(): > 0) E(X) = Var(X) = , Pr(X = j) = e , j {0, 1, . . . }, j! Pr(X = j) = (1 q)q j , j {0, 1, . . . },
j

G(z) = e(1z) .

Geometric (0 < q < 1) E(X) = q/(1 q), Var(X) = q/(1 q)2 , Note: Pr(X j) = q j G(z) =
n 1q 1qz .

Negative binomial (0 < q < 1, n 1) E(X) = nq/(1 q), Var(X) = nq/(1 q)2 , Pr(X = j) =
n+j1 j

(1 q)n q j , j {0, 1, . . . },

G(z) =

1q 1qz

Hypergeometric (N 0, 0 n, a N ) E(X) = na/N , Var(X) = na(N n)(N a)/(N 2 (N 1)), Pr(X = j) =

, j {max(0, n + a N ), . . . , min(n, a)},

G(z) = complicated.

21. Continuous distributions: Here X is a continuous rv taking values in a subset of R. The mean, variance, pdf f : R [0, ) and (if it can be written down explicitly) the distribution function F : R [0, 1] are listed; f takes the value 0 outside the range given, so that F takes the value 0 below that range and 1 above. The mgf M (t) = E(etX ), or cf (t) = E(eitX ), whichever is appropriate, is also listed. For non-negative rvs, the LST satises L(t) = E(etX ) = M (t), t 0. Uniform (U (a, b): a < b) E(X) = (a + b)/2, Var(X) = (b a)2 /12, f (x) =
1 ba

, F (x) =

xa ba

, a x b,

M (0) = 1, M (t) =

ebt eat (ba)t ,

t = 0.

Exponential (exp(): > 0) E(X) = 1/, Var(X) = 1/2 , f (x) = ex , F (x) = 1 ex , x 0, M (t) =
2 t

, t < .

Gamma ((, ): > 0, > 0) E(X) = /, Var(X) = / , f (x) = ex (x) () , x 0,

M (t) =

, t < .

The Chi-squared distribution 2 (n 1) is (n/2, 1/2). The Erlang distribution is (n, ), and n
x f (x) = ex (x) (n1)! , F (x) = 1 e
n1

n1 (x)i i=0 i!

, x 0.

Beta (a > 0, b > 0) E(X) = a/(a + b), Var(X) = ab/((a + b)2 (a + b + 1)), f (x) =
1 a1 (1 (a,b) x

x)b1 , 0 x 1,

M (t) = complicated.

Normal (Gaussian) (N (, 2 ): R, 2 > 0) E(X) = , Var(X) = 2 , f (x) =

exp 1 (x )2 / 2 , x R, 2

M (t) = exp t + 1 2 t2 , t R. 2 M (t) = exp tT + 1 tV tT , t Rn . 2

Multivariate Normal (N (, V ): Rn , V +ve-denite symmetric) E(X) = , Cov(X) = V , f (x) =

1 (x )V 1 (x )T , x Rn , 2

Cauchy (m R, b > 0) median= m (Note that E(X) does not exist: E(X + ) = E(X ) = ) f (x) =
b (b2 +(xm)2 ) ,

F (x) =

1 2

arctan

xm b

, x R,

(t) = eimtb|t| , t R.

Weibull ( > 0, > 0) E(X) = 1/ (1 + 1/), Var(X) = 2/ {(1 + 2/) ((1 + 1/))2 }, f (x) = x1 exp x , F (x) = 1 exp x , x 0, Laplace ( R, > 0) E(X) = , Var(X) = 2 , f (x) =
1 2 2 et 1 2 t2 ,

M (t) = complicated.

exp (|x |/) , x R,

M (t) =

|t| < 1/.