Robert Martin

HL Maths notes
1 Algebra
1.1 Sequences and Series

Arithmetic progressions

• Tn = Un = a + (n − 1)d.
• A sequence is an A.P if Tn − Tn−1 = d = constant.
n n
• Sn = 2 (a + l) = 2 (2a + (n − 1)d).
• Tn = Sn − Sn−1 .

Geometric progressions

• Tn = arn−1 .
Tn
• A sequence is a G.P if Tn−1 = r = constant.
a(1−r n )
• Sn = 1−r .
a
• |r| < 1 =⇒ S∞ = 1−r .

• |r| > 1 =⇒ divergent.

1.2 Summation
n
P
For ur , the number of terms is (n − m + 1).
r=m
n
X n
X n
X
(xr ± yr ) = xr ± yr
r=1 r=1 r=1
n
X n
X
kur = k ur
r=1 r=1
n
X n
X m−1
X
ur = ur − ur
r=m r=1 r=1

Useful sums:
n
X 1
r= n(n + 1)
r=1
2
n
X 1
r2 = n(n + 1)(2n + 1)
r=1
6
n n
X X 1
r3 = ( r)2 = n2 (n + 1)2
r=1 r=1
4

1
 Robert Martin

1.3 Permutations and combinations

n n! n
Cr = Pr =n Cr · r!
r!(n − r)!
n!
If m objects are identical and the remaining are distinct (a total of n objects), permutations = m!

1.4 The Binomial Thoerem

n(n − 1) 2 n(n − 1)(n − 2) 3
(1 + x)n = 1 + nx + x + x + ···
2! 3!

1.5 Mathematical induction

1. Let Pn be the statement: ello for all n ∈ Z+ .
2. For n = 1: LHS = something. RHS = something =⇒ P1 is true.
3. Assume Pk is true for some k ∈ Z+ .

4. Showing that Pk+1 is true: it is true!

5. Since P1 is true, and Pk is true =⇒ Pk+1 is true, by Mathematical Induction, Pn is true for all n ∈ Z+ .

To do the inductive step:

k+1
X Xk
• ur = uk+1 + ur
r=1 r=1
k+1 k
d y d d y
• = ( )
dxk+1 dx dxk
• For divisibility, let the expression = a multiple of m. You can always rearrange the inductive hypothesis.

2 Functions and equations

• A function is a to-one relationship.

• If the vertical line x = a cuts the graph at one point only, then f is a function. If it cuts more than once,
give an example.
• If a function passes the horizontal line test, it will have an inverse.
• The inverse is just a reflection of the graph in the line y = x.

• For inverse functions, Rf = Df −1 and Df = Rf −1 .

• For gf to exist, Rf ⊆ Dg .
• Dgf = Df .

• Rgf = Rg |(Dg = Rf ).
• (g ◦ f )−1 (x) = (f −1 ◦ g −1 )(x).
• f f −1 (x) may not necessarily intersect with f −1 f (x), it depends on the domain.

• For a periodic function, f (x) = f (x + c).

2
 Robert Martin

2.1 Graphs
• To transform, TSST. (translate and stretch)x then (translate and stretch)y.

• For y = |f (x)|, retain y ≥ 0, then reflect y < 0.

• For y = f (|x|), retain x ≥ 0, then reflect x ≥ 0 to the left of the x-axis.
• For each transformation, you’re allowed to replace x by something else.

2.2 Polynomials
• For a polynomial of degree n:
– The sum of individual roots = − an−1
an
an−2
– The sum of (choose 2) roots = an
– The sum of (choose 3) roots = − an−3
an
– The product of roots, i.e the sum of (choose n) roots = (−1)n aano

• For the special case of a quadratic: α + β = − ab , αβ = c

a

• A polynomial of degree n has a maximum of n roots, but some of these may be complex.

2.3 Circular functions and Trigonometry

• The ambiguous case of the sine rule occurs when the angle you are trying to find is opposite the longest
side.
• sin(−θ) = − sin θ tan(−θ) = − tan θ (odd functions).
• cos(−θ) = cos θ (even function).

• For π ± θ or 2π ± θ: sin-sin, cos-cos, tan-tan.

π 3π
• For 2 ± θ or 2 ± θ: sin-cos, cos-sin, tan-cot.
• tan x = cot( π2 − x).
• sec x = csc( π2 − x).

• The domain of arcsin x and arccos x are [−1,1].

√
• cos(arcsin x) = sin(arccos x) = 1 − x2 .
• A circle with centre (h, k) and radius r is described by:

(x − h)2 + (y − k)2 = r2

• To simplify an expression with trig, it may help to use the half angle formula.

sin θ 2 sin θ2 cos θ2 θ

= θ
= tan
1 + cos θ 2
1 + 2 cos 2 − 1 2

3
 Robert Martin

2.4 Systems of equations

• A system of equations can be written as an augmented matrix:
 
2x + 3y + 4z = 2 2 3 4 2
3x − 2y + z = −3 → 3 −2 1 −3
x + 4y − z = 5 1 4 −1 5

• A system is consistent if it has solutions.

• A system is inconsistent if one of the rows reduces to 0 = a.
• If the last row reduces to 0 = 0, there are infinitely many solutions and the general solution can be found
by setting z = λ where λ is a real parameter.

• If the determinant of the 3 × 3 matrix is zero, then there is no unique solution (i.e either no solutions or
infinite solutions).
• This links to planes, since the Cartesian equation of a plane is ax + by + cz = d.

3 Vectors
−−→
• A vector AB can be represented by a straight line, with an arrow, joining A and B.
• A vector can also be denoted with a lower case letter, e.g a, which is written with a tilde below it.
−→
• A position vector defines the position of a point relative to the origin. a = OA.
 
x
• The Cartesian form of a vector: r = xi + yj + zk, or r = y 
z
p
• |r| = x2 + y 2 + z 2
r
• A unit vector: r̂ = |r|

−−→ −
−→ −−
→
µOA+λOB
• The Ratio Thoerem: OP = µ+λ

3.1 Scalar products

• The scalar product of two vectors is defined as a · b = |a||b| cos θ.
• The vectors must both converge or diverge from one point.

• a · a = |a|2 .
• Most algebra works, except for cancellation and division.

4
 Robert Martin

• a⊥b ⇐⇒ a · b = 0.
   
a1 b1
• a2  · b2  = a1 b1 + a2 b2 + a3 b3 .
a3 b3

3.2 Vector products

• The vector product of two vectors is defined as a × b = |a||b| sin θn̂.
• n̂ is a unit vector perpendicular to a and b.
• (a × b) = −(b × a).

• (λa) × (µb) = (λµ)(a × b).

• |a × b| = |a||b| sin θ.
• a k b ⇐⇒ a × b = 0, hence a × a = 0.
• a⊥b ⇐⇒ a × b = |a||b|.

• a · (a × b) = 0.
     
a1 b1 a2 b3 − b2 a3
• a2  × b2  = −(a1 b3 − b1 a3 ). Cover top find det, cover mid find negative det, cover bot find det.
a3 b3 a1 b2 − b1 a2
−−→ −→ −−→ −→
• Area ∆ABC = 21 |AB||AC| sin θ = 12 |AB × AC|.

3.3 Projections and resolving vectors

−−→
• The length of the horizontal projection of a onto b = ON = |a||b̂| cos θ = a · b̂
• The length of the vertical projection is given by |AN | = |a × b̂|
• The horizontal projection vector is then u = (a · b̂)b̂, which is the same as the resolved component of a
parallel to b.

• The perpendicular component of a is v = a − u.

3.4 Straight lines

l : r = a + λd, λ ∈ R.

• The vector equation of a line uses a position vector a of a fixed point on l, and a direction vector d parallel
to l, to find the position vector of any point on the line (r).

• λ is a real parameter, which means that the vector equation of a line is not unique.

5
 Robert Martin

• To get the parametric form, we write the equation as column vectors then equate components:

x = a1 + λd1 ,

y = a2 + λd2 , λ ∈ R

z = a3 + λd3 ,


• To get the Cartesian form, make λ the subject then eliminate it.
 
x−a1
x = a1 + λd1 ,
  d1 = λ,

x − a1 y − a2 z − a3
y−a2
y = a2 + λd2 , =⇒ d2 = λ,
=⇒ = = (= λ)
  z−a3
 d1 d2 d3
z = a3 + λd3 = λ

d3

• l1 and l2 are parallel ⇐⇒ d1 and d2 are parallel ⇐⇒ d1 = kd2 , for some k ∈ R.

• l1 and l2 intersect ⇐⇒
– d1 is not parallel to d2 AND
– there exist unique values of λ and µ such that a1 + λd1 = a2 + µd2 .

• The lines are skew ⇐⇒ the direction vectors aren’t parallel and there aren’t unique values of λ and µ.
·d2
• The acute angle between two lines is given by cos−1 | |dd11||d2|
|.

3.5 Planes

−→ −→
AP ⊥n =⇒ AP · n = 0 =⇒ (r − a) · n = 0.
• The scalar product form of the vector equation of the plane is r · n = a · n, where a is a fixed point on
the plane.
• n can be found by taking the cross product of two known vectors parallel to the plane.
• The shortest distance between the origin and the plane: |d| = |a · n̂|
• The parametric form of the vector equation of the plane:

Π : r = a + λd1 + µd2 , λ, µ ∈ R

• By expanding the scalar product form, we can arrive at the Cartesian form:
   
x a
r · n = y  ·  b  = ax + by + cz = D
z c

• A line will be parallel to a plane if it is perpendicular to n, i.e n · d = 0 and there is no common point.

6
 Robert Martin

• If not parallel, it will intersect at a point, which can be found by substituting the line equation into the
plane equation.
d·n
• The acute angle between l and Π: sin θ = | |d||n| |

• When planes intersect, their Cartesian forms can be combined to form a system of simultaneous equations
– If there is a unique solution, the planes intersect at a point.
– If there are infinitely many solutions, the planes intersect in a line.
– If there are no solutions, the three planes do not intersect.

4 Calculus
4.1 Differentiation
• If the limit of the denominator of a rational function is zero, you cannot substitute to find the limit: either
‘juggle’ or use l’Hopital’s rule, e.g:
 
sin x  cos x 
lim = lim =1
x→0 x x→0 1

• The definition of the derivative:

 
f (x + δx) − f (x)
f 0 (x) = lim
δx→0 δx

• Special derivatives:
d
dx (sec x) = sec x tan x
d
dx (csc x) = − csc x cot x
d
dx (cot x) = − csc2 x
d √ 1
dx (arcsin x) = 1−x2

d 1
dx (arccos x) = − √1−x 2

d 1
dx (arctan x) = 1+x2

d x
dx (a ) = ax ln a
d 1
dx (loga x) = x ln a
dy
• f (x) is an increasing function on (a, b) if dx ≥ 0 on that interval, or a strictly increasing function
dy
if dx > 0.
d2 y
• f (x) is concave upwards on (a, b) if dx2 > 0.
• If the derivative at a point is zero, the function is stationary.
• If the derivative at a point is ∞, there is a vertical line.
d2 y d2 y
• For a point of inflexion, dx2 = 0 AND the sign of dx2 changes, i.e concativity changes.

7
 Robert Martin

• Sketching the graph of f 0 (x) given f (x):

– Stationary point → x-intercept.
– f (x) increasing → f 0 (x) above x-axis.
– Point of inflexion → turning point.
dy
• The gradient at any point on the curve: m = dx |x=x0 .

• The equation of a tangent to the curve at (x0 , y0 ): y − y0 = m(x − x0 ).

• If two variabels are related, their rates of change are also related:
dy dy dx
= ·
dt dx dt

• In kinematics especially:
dv dv ds dv
a= = · =v
dt dx dt ds

4.2 Integration
(px + q)n+1
Z
(px + q)n dx = +C
p(n + 1)

(f (x))n+1
Z
f 0 (x)(f (x))n dx = +C
n+1

f 0 (x)
Z
dx = ln |f (x)| + C
f (x)

x−1
Z Z
1
dx = + C = ln | ln |x|| + C
x ln x ln x
Z
tan x dx = ln | sec x| + C
Z
sec x dx = ln | sec x + tan x| + C
Z
csc x dx = − ln | csc x + cot x| + C
Z
1 1 x+k
dx = arctan( )+C
(x + k)2 + a2 a a
Z
1 x+k
p dx = arcsin( )+C
a2 − (x + k)2 a

• To integrate sin2 x or cos2 x, we expand cos(2x) and rearrange.

• To integrate sin3 x, split into sin x(sin2 x)dx, then use sin2 x + cos2 x = 1.
R

• If the integral is of the form:

Z Z
px + q px + q
√ dx or 2 + Bx + C
dx
2
Ax + Bx + C Ax
R 0 (x)
use sorcery to change it into ff (x) dx or f 0 (x)(f (x))n dx.
R

8
 Robert Martin

• Integration by substitution:
1. Replace dx by dx
dt · dt.
2. Substitute by replacing all x with g(t).
Then: f (x) dx = f (g(t)) dx
R R
dt · dt

• Integration by parts:
R R
udv = uv − vdu
• To choose which one to differentiate, use LIATE: Logs, Inverse trig, Algebraic, Trig, Exponentials.

4.3 Definite integrals

Rb Ra
• a
f (x)dx = −f (x)dx
b
Rb
• The definite integral a f (x)dx can only be found if f (x) is defined for all x ∈ (a, b).
Rb
• The area between a curve and the y-axis: a f (y)dy
• If a function is difficult to integrate, try integrating its inverse w.r.t y then subtract from a rectangle. e.g:

Z √
3
√ Z π3
π 3
arctan x dx = − tan y dy
0 3 0

Rb
• The area between the curve and the axis is always a |f (x)|dx.
Rb
• The area between two curves is always a y1 − y2 dx.
• The volume of revolution: Z b
V =π y 2 dx
a

• The volume of revolution of the area enclosed by two curves:

Z b Z b
V =π (y1 )2 dx − π (y2 )2 dx
a a

5 Probability and Statistics

5.1 Probability
• Two events A and B are mutually exclusive if P (A ∩ B) = 0.
• P (A ∪ B) = P (A) + P (B) − P (A ∩ B).
P (A∩B)
• P (A|B) = P (B) .

• A and B are independent if P (A|B) = P (A), so if they are independent P (A ∩ B) = P (A)P (B).

9
 Robert Martin

5.2 Discrete random variables

• P (X = x) is the probability that the r.v X will assume a value of x.

• A discrete r.v can assume a countable number of values.

• For a d.r.v taking values x1 , x2 , x3 , ..., xn , the probability distribution is defined as P (X = xi ), such
that: X
0 ≤ P (X = xi ) ≤ 1 and P (X = xi ) = 1
all i

• The expectation of a d.r.v:

X
E(X) = µ = xP (X = x)
X
E(g(X)) = g(x)P (X = x)
E(a) = a
E(aX ± b) = aE(X) ± b
E(X ± Y ) = E(X) ± E(Y )
• The variance of a d.r.v:
Var(X) = σ 2 = E((x − µ)2 ) = E(X 2 ) − [E(X)]2
Var(a) = 0
Var(aX + b) = a2 Var(X)
Var(X ± Y ) = Var(X) + Var(Y ) (only if X and Y are independent)
• Note: never subtract variance.

5.3 Discrete distributions

The Binomial distribution

 
n x n−x
X ∼ B(n, p) P (X = x) = p q E(X) = np Var(X) = npq
x

• There are n independent trials, two possible outcomes (either ‘success’ or ‘failure’), with constant proba-
bility of success p, X is the number of ‘successes’.
• The Binomial distribution is a combination of n Bernoulli trials.
• For P (X ≤ x), we find P (X = 0) + P (X = 1) + P (X = 2) + ... + P (X = x).

The Poisson distribution

e−λ λx
X ∼ P o(λ) P (X = x) = E(X) = Var(X) = λ
x!
• For a random variable in time or space, if there is no chance of simultaneous events, the events are
independent, and the events have a constant probability of occuring, it is a Poisson process.
• λ is the parameter, and defines the number of events in a given time/space.
• If X ∼ P o(λ) and Y ∼ P o(µ), then X + Y ∼ P o(λ + µ).

10
 Robert Martin

The Geometric distribution

1 q
X ∼ Geo(p) P (X = x) = pq x−1 , x ≥ 1 E(X) = Var(X) =
p p2
If we perform a series of independent trials with a probability p of success, X is the number of trials up
to and including the first success.
P (X > x) = P (X = x + 1) + P (X = x + 2) + . . .
= pq x + pq x+1 + pq x+2 + . . .
1
= pq x (1 + q + q 2 + ...) = pq x ( ) = qx
1−q
P (X > a + b|X > a) = P (X > b) = q b

The Negative Binomial distribution

 
x − 1 r x−r r rq
X ∼ N B(r, p) P (X = x) = p q , r ≥ 1, x ≥ 1 E(X) = Var(X) =
r−1 p p2
• X is the number of trials needed to achieve r successes.
• The Negative Binomial distribution is just a combination of r geometric trials.

5.4 Continuous random variables and CDFs

• Instead of probability distributions, we have probability density functions (PDFs), denoted by f (x).
– f (x) ≥ 0 for all x ∈ R
R∞
– −∞ f (x) dx = 1
• Continuous =⇒ uncountable, so P (X = x) = 0. Therefore, ≥ or > is irrelevant.
Z b
P (a < X < b) = f (x) dx
Z ∞ a
E(X) = µ = xf (x) dx
−∞
Z ∞
E(g(X)) = g(x)f (x) dx
−∞
P (|X − a| < b) = P (−b < X − a < b)
• The mode of a c.r.v is the value of x which gives the maximum probability, i.e the x coordinate of the
highest point in the domain.
• The cumulative distribution function (CDF):
Z x
F (x) = P (X ≤ x) = f (t) dt
−∞
lim F (x) = 0 lim F (x) = 1
x→−∞ x→∞

P (a < X < b) = F (b) − F (a)

d
F (x) = f (x)
dx
• F (x) is continuous and increasing (since f (x) > 0).
1
Rm
• To find the median m, set F (m) = 2 and solve for m, i.e: −∞
f (t)dt = 0.5

11
 Robert Martin

5.5 The Normal distribution

X ∼ N (µ, σ 2 )

• The Normal distribution is a bell curve symmetrical about x = µ.

• The mean = median = mode = µ.
• µ affects the location of the curve, whereas σ 2 affects the spread.
• The standard normal distribution is denoted by Z ∼ N (0, 1).
X−µ
• Any normal distribution can be standardised: Z = σ

• The Z score represents the number of standard deviations away from the mean.
• To find c given P (X < c) = p, use invNorm.

• If X ∼ N (µ1 , σ12 ) and Y ∼ N (µ2 , σ22 ), then aX + bY also has a normal distribution.

E(aX + bY ) = aE(X) + bE(Y )

= aµ1 + bµ2
Var(aX + bY ) = a2 σ12 + b2 σ22

aX + bY ∼ N (aµ1 + bµ2 , a2 σ12 + b2 σ22 )

5.6 Sampling
• If X is a random variable, X1 , X2 , X3 , ..., Xn are a sample of n independent observations.
• The sample mean:
X1 + X2 + ... + Xn
X̄ =
n
 
X1 + X2 + ... + Xn nE(X)
E(X̄) = E = = E(X) = µ
n n

σ2
 

X1 + X2 + ... + Xn 1 nVar(X)
Var X̄ = Var = 2 Var(X1 + X2 + ... + Xn ) = 2
=
n n n n

• For the sample sum: E(S) = nµ, Var(S) = nσ 2

• Therefore, in a normal population:
n
σ2 X
X̄ ∼ N (µ, ) Xr ∼ N (nµ, nσ 2 )
n r=1

• The Central Limit Theorem states that, for a large sample size (n ≥ 50), the sample mean/sum of a
sample from any distribution (e.g not normal), will approximately follow the normal distribution.

12
 Robert Martin

5.7 Estimators
• An estimator is a test statistic T based on observed data that estimates an unknown parameter θ.
• The estimator is unbiased if E(T ) = θ.
• The sample mean is an unbiased estimator of µ since E(X̄) = µ.
n−1 2
• However, the sample variance is not an unbiased estimator for σ 2 since E(Sn2 ) = n σ .

• An unbiased estimator for σ 2 :

 X 
n n 1
s2n−1 = × Sn2 = x2 − (x̄)2
n−1 n−1 n
( x)2
X P 
1
= x2 −
n−1 n

• An unbiased estimator is more efficient than another if it has a lower variance.

5.8 Confidence intervals

• A 95% confidence interval (CI) means that there is a 95% chance that the interval includes µ.
• For X ∼ N (µ, σ 2 ), if we take a sample: X̄ ∼ N (µ, σ 2 ).
σ
Confidence limits = X̄ ± Zk √
n
 
σ σ
CI = X̄ − Zk √ , X̄ + Zk √
n n

• Zk is the critical value, and is found using invNorm.

• For a 95% CI: invNorm(0.025) = −1.96

• The width of a CI is 2Zk √σn

• If we have a large sample from any population (µ and σ 2 unknown), we can use the CLT.
 
sn−1 sn−1
CI = x̄ − Zk √ , x̄ + Zk √
n n

• If the population is normal but we do not know the variance, we use the t-distribution.
X̄ − µ
T = sn−1/√n
follows a t-distribution with n − 1 degrees of freedom.

 
sn−1 sn−1
CI = x̄ − tk √ , x̄ + tk √
n n

13
 Robert Martin

σ2 n Assumptions Test Statistic

large CLT X̄−µ
known Z= σ/√n ∼ N (0, 1)
small normal
X̄−µ
large CLT Z= sn−1/√n ∼ N (0, 1)
unknown X̄−µ
small normal T = sn−1/√n ∼ tn−1

5.9 Hypothesis testing

1. State H0 and H1 .
2. Test statistic.
3. Level of significance and rejection criteria.

4. Compute p-value (or z-value or t-value).

5. Conclusion in context.

e.g

H0 : µ = 3
H1 : µ > 3
X̄ − µ
Test statistic: Z = σ/√n

Sig level = 5%, one tailed.

Reject H0 if p < 0.05
Since p-value = 0.03 < 0.05, we reject H0 and conclude that there is significant evidence at the 5% level that...

• P (Type I Error) = P (H0 rejected|H0 true) = α%. i.e P (Type I Error) = significance level.
• P (Type II Error) = P (H0 accepted|H1 true).
• For example, for H0 : µ = µ0 H1 : µ = µ1 ,

P (Type II Error) = P (H0 accepted|H1 true) = P (X̄ < critical value |X̄ ∼ N (µ1 , σ 2 ))

14
 Robert Martin

5.10 PGFs
X
G(t) = E(tX ) = tx P (X = x)
G(1) = 1
X
G0 (t) = xtx−1 P (X = x) ∴ E(X) = G0 (1)

X
G00 (t) = x(x − 1)tx−2 P (X = x)
X X
G00 (1) = x2 P (X = x) − xP (X = x) = E(X 2 ) − E(X)
∴ E(X 2 ) = G00 (1) + G0 (1)
∴ Var(X) = G00 (1) + G0 (1) − [G0 (1)]2

If Z = X + Y, GZ (t) = E(tZ ) = E(tX+Y ) = E(tX )E(tY ) = GX (t)GY (t)

G(n) (0)
• To find P (X = n), we use the Maclaurin series: P (X = n) = n! .
• To prove most things about PGFs, differentiation will be involved (sometimes using the product rule and
chain rule).
Binomial

If Y ∼ B(n, p), we can say that Y = X1 + X2 + X3 + ... + Xn where X is a Bernoulli trial.

x 0 1
P (X = x) q p
X
GX (t) = tx P (X = x) = q + pt
n n
GY (t) = E(tY ) = E(tX1 +...+Xn ) = E(tX ) = [GX (t)] = (q + pt)n


Poisson

e−λ λx
If X ∼ P o(λ), P (X = x) = x! .
X
G(t) = E(tX ) = tx P (X = x)
X e−λ λx
= tx
x!
X (λt)x
= e−λ = e−λ eλt = eλ(t−1) .
x!
Geometric

If X ∼ Geo(p), P (X = x) = pq x−1 .

X
G(t) = E(tX ) = tx P (X = x)
X
= tx pq x−1
= pt + pt2 q + pt3 q 2 + pt4 q 3 + ... + ptn q n−1 + ...
a pt
S∞ = =
1−r 1 − qt

15
 Robert Martin

Negative Binomial

If Y ∼ N B(r, p), we can say that Y = X1 + X2 + X3 + ... + Xr , where X ∼ Geo(p).

 r
Y X1 +...+Xr
 X r
 r pt
GY (t) = E(t ) = E(t ) = E(t ) = [GX (t)] =
1 − qt

5.11 Bivariate data and correlations

• If X and Y are random variables, the joint probability distribution is P (X = x ∩ Y = y).
PP
• p(x, y) = 1
PP
• E(XY ) = xy p(x, y)
• Cov(X, Y ) = E(XY ) − E(X)E(Y ). X and Y independent =⇒ Cov(X, Y ) = 0.
• Var(X + Y ) = Var(X) + Var(Y ) − 2Cov(X, Y ).
• The correlation coefficient measures the linear relationship between X and Y
Cov(X, Y ) σXY
ρ = Corr(X, Y ) = =
σX σY σX σY

• A bivariate sample consists of pairs of data (x1 , y1 ). For a bivariate sample, the above points do not
apply.
( x)2
P P
xy − nx̄ȳ Sxy X
2
r= pP =p , where Sxx = x −
n
P
( x2 − nx̄2 )( y 2 − nȳ 2 ) Sxx Syy
• If r = 0, there is no linear relationship, but it does not imply that X and Y are independent.
• r is independent of the units, and does not show any causality.
• In maths, controlled variable = independent variable.
• The y-on-x regression line y = a + bx will always pass through (x̄, ȳ).
Sxy
y − ȳ = b(x − x̄), where b =
Sxx
• The x-on-y regression line is denoted by x = c + dy.
√
bd = r2 r = ± bd, the sign depends on whether the gradient is positive or negative.

• We can statistically test evidence of a correlation by assuming both variables follow a bivariate normal
distribution with correlation coefficient ρ:
H0 : ρ = 0
H1 : ρ 6= 0
r
n−2
Test statistic: T = r ∼ tn−2
1 − r2
Sig level = 5%, two tailed.
Reject H0 if |T | > invt(0.975, n − 2)
r
n−2
Note: T = r (sub in values)
1 − r2
Since |T | = 0.08 > invt(0.975, n − 2), we reject H0 and conclude that there is significant evidence at the
5% level that there is a correlation between...

16
 Robert Martin

6 Complex numbers
6.1 Forms of complex numbers
• The Cartesian form of a complex number: z = x + iy. This relates a complex number to its real and
imaginary parts. x = Re(z), y = Im(z).
• The Polar form, a.k.a the trigonometric form or modulus-argument form:

z = r(cos θ + i sin θ) = rcis(θ)

p
• r is the modulus of z: r = |z| = x2 + y 2 .
−→
• The argument of z (θ or arg z) is the angle from the positive real axis to the line OZ. The principal
value of arg z is the angle in the interval (−π, π].
– The argument can be found using arctan(y/x), but you must consider the quadrant.
– arg 2 = 0 arg (−3) = π
– arg (3i) = π/2 arg (−4i) = −π/2
– arg 0 is undefined.
• Using the Maclaurin expansions of ex , cos x and sin x, we can derive Euler’s beautiful formula:

eix = cos x + i sin x

• We can then write complex numbers in the exponential or Euler form: z = reiθ , for θ in radians.

Complex conjugates

• The conjugate of z is given by z ∗ = x − iy.

• It is interpreted on an Argand diagram as a reflection in the real axis.
• Because of this, arg z = − arg z ∗ so z ∗ = rcis(−θ) = re−iθ .
• Properties of conjugates

– (z ∗ )∗ = z
– (z + w)∗ = z ∗ + w∗
– (zw)∗ = z ∗ w∗ =⇒ (z n )∗ = (z ∗ )n
– z + z ∗ = 2Re(z)
– z − z ∗ = 2iIm(z)
– zz ∗ = x2 + y 2 = |z|2
– z ∗ = r2 /z

17
 Robert Martin

6.2 Operations on complex numbers

• When adding and subtracting complex numbers, we group real and imaginary parts.
• To multiply complex numbers in Cartesian form, we can expand the brackets.
• To multiply complex numbers in the Euler form, multiply moduli and add arguments:
z1 z2 = (r1 eiθ1 )(r2 eiθ2 ) = r1 r2 ei(θ1 +θ2 )

• To divide complex numbers, we subtract their arguments.

• De Moivre’s Theorem states that, if z = r(cos θ + i sin θ),
z n = rn (cos nθ + i sin nθ) , for all n ∈ R

• It follows that |z n | = |z|n .

6.3 Relation to trigonometry

z + z ∗ = eiθ + e−iθ = (cos θ + i sin θ) + (cos θ − i sin θ) = 2 cos θ
z − z ∗ = eiθ − e−iθ = (cos θ + i sin θ) − (cos θ − i sin θ) = 2i sin θ
eiθ + e−iθ eiθ − e−iθ
=⇒ cos θ = and sin θ =
2 2i
When simplifying expressions involving eiθ ± 1, we can use this trick:
θ θ θ θ θ
eiθ + 1 = ei 2 (ei 2 + e−i 2 ) = 2ei 2 cos
2
θ θ θ θ θ
eiθ − 1 = ei 2 (ei 2 − e−i 2 ) = 2iei 2 sin
2
Trigonometric identities

• Write cos 3θ in terms of cos θ.

3
cos 3θ = Re(cos 3θ + i sin 3θ) = Re((cos θ + i sin θ) ) (by De Moivre’s Theorem).

But using a binomial expansion, (cos θ + i sin θ)3 = cos3 θ + 3 cos2 θ(i sin θ) + 3 cos θ(i sin θ)2 + (i sin θ)3

cos 3θ = Re(cos3 θ + 3 cos2 θ(i sin θ) + 3 cos θ(i sin θ)2 + (i sin θ)3 )
=⇒ cos 3θ = cos3 θ + 3 cos θ(i sin θ)2 = cos3 θ − cos θ(1 − cos2 θ)

∴ cos 3θ = 4 cos3 θ − cos θ. QED.

• Express sin3 θ in terms of sines of multiples of θ. To begin, let z = cis(θ).
3
3z 2
    
1 3z 1 1 1
z− = z3 − + 2 − 3 = z3 − 3 − 3 z −
z z z z z z
 
1
For a complex number of unit modulus, z n − n = (z n − (z n )∗ ) = 2i sin nθ
z
=⇒ (2i sin θ)3 = 2i sin 3θ − 3(2i sin θ)
=⇒ −8i sin3 θ = 2i sin 3θ − 6i sin θ
3 1
∴ sin3 θ = sin θ − sin 3θ. QED.
4 4
• For cosines, we instead use z + z1 .

18
 Robert Martin

6.4 Polynomials
• A quadratic will have complex roots if the discriminant b2 − 4ac < 0.

• In general, the complex roots of a quadratic with real coefficients will always be a conjugate pair.
• A cubic will either have 3 real roots or 1 real root and 2 conjugate complex roots. If we know one of the
complex roots, we know its conjugate and can multiply out. Long division will help us find the real root.

(x − (a + bi))(x − (a − bi)) = x2 − 2ax + (a2 + b2 )

(x − z)(x − z ∗ ) = x2 − 2Re(z) + |z|2

6.5 Roots of complex numbers

• There are n values of z that solve z n = 1 (because of the Fundamental Theorem of Algebra); these are
known as the nth roots of unity.
• To find these, we rewrite the RHS: 1 = ei(0+2kπ) . As a result,
2kπ
z = ei n , for k = 1, 2, 3, ..., n.

• Alternatively, use k = 0, ±1, ±2, ... in order to make sure that arguments will be within the principal
range.
• Note that each of the roots will form on a unit circle.

• More generally, for the nth roots of a complex number c,

θ+2kπ
z = r1/n ei n , for k = 1, 2, 3, ..., n.

7 Miscellaneous
• (a ± b)3 = a3 ± 3a2 b + 3ab2 ± b3
• a3 + b3 = (a + b)(a2 − ab + b2 )

• a3 − b3 = (a − b)(a2 + ab + b2 )
• |x + 3||x + 2| = |(x + 3)(x + 2)|

19