Applied Statistics Concise Exam Notes

Applied Statistics
Concise Exam Notes: distributions, estimation, tests, chi-square, F and
regression
Clean formulas + when to use + worked patterns

1. Basic Language of Statistics

Population is the whole group; sample is the part observed. A parameter describes the population,
while a statistic is calculated from the sample.

Quantity Meaning Symbol

Population mean true average, usually unknown µ
∑x
Sample mean average from data x̄ =
n
Population SD true spread σ
Sample SD estimated spread s
Variance squared spread σ 2 , s2

Mean and spread

∑x ∑( x − x̄ )2 √
x̄ =, s2 = , s = s2 .
n n−1
Why divide by n − 1? A sample has already used one piece of information to find x̄, so only n − 1
independent deviations remain.

2. Random Variables and Probability Distributions

A random variable assigns a number to a random outcome. A discrete variable counts values; a con-
tinuous variable measures values.

Type Typical question Main distribution

Fixed number of success/failure trials “exactly 3 defective out of 10” Binomial
Events per interval “calls per hour”, “earthquakes Poisson
per month”
Measurement with bell-shaped pattern lifetime, weight, diameter Normal
Waiting time until the rth event time until 3rd arrival Gamma
Lifetime/reliability/failure survives beyond time t Weibull

2.1. Binomial distribution

Use when there are n independent trials, each trial has success/failure, and the success probability is
constant p.

Binomial: X ∼ Bin(n, p)
 
n x
P( X = x ) = p (1 − p ) n − x , E( X ) = np, Var( X ) = np(1 − p).
x

1
Applied Statistics Concise Exam Notes

Worked pattern

A piston is rejected with probability 0.15. For 10 pistons, find exactly 2 rejects.
 
10
P ( X = 2) = (0.15)2 (0.85)8 .
2

Why this formula? p2 (1 − p)8 gives one exact arrangement, and (10
2 ) counts all positions of the two
rejects.

Exam wording

exactly 2 = P( X = 2), at most 2 = P( X ≤ 2), at least 2 = P( X ≥ 2) = 1 − P( X ≤ 1).

2.2. Poisson distribution

Use when counting random independent events in a fixed time, area or length, with known average
rate λ.

Poisson: X ∼ Poisson(λ)

e−λ λ x
P( X = x ) = , E( X ) = λ, Var( X ) = λ.
x!

Worked pattern

Earthquakes occur on average 0.77 per month. Probability of exactly 2 in a month:

e−0.77 (0.77)2
P ( X = 2) = .
2!
Probability of at least one:

P( X ≥ 1) = 1 − P( X = 0) = 1 − e−0.77 .

3. Continuous Distributions
3.1. Normal distribution
The normal distribution is symmetric and bell-shaped. Most exam questions reduce to the standard
normal table.
Standard normal curve

P ( Z < 1)

−3 −2 −1 1 2 3

2
Applied Statistics Concise Exam Notes

Standardization
If X ∼ N (µ, σ2 ), then
X−µ
Z= ∼ N (0, 1).
σ
For a sample mean,
σ2 X̄ − µ
 
X̄ ∼ N µ, , Z= √ .
n σ/ n

√ by σ? It changes the distance from the mean into number of standard deviations. Why
Why divide
divide by n for X̄? Averages vary less than individual observations.

Worked pattern

Battery lifetime is normal with µ = 1200 and σ = 150. Find P( X < 1350).

1350 − 1200
Z= = 1, P( X < 1350) = P( Z < 1) = 0.8413.
150

3.2. Gamma and Weibull distributions

Distribution Use Core formula

Gamma waiting time until the αth event in a E( X ) = α/λ, Var( X ) = α/λ2
Poisson process
F ( x ) = 1 − e−(x/β) , P( X > x ) =
α
Weibull lifetime, survival, reliability and failure
e−(x/β)
α
time

How to identify them quickly

Poisson counts events; Gamma measures waiting time until an event count is reached. Weibull is
used when the question says failure, lifetime, survival or reliability.

4. Confidence Intervals
A confidence interval gives a plausible range for an unknown parameter.

General structure

estimate ± critical value × standard error.

Situation Confidence interval for µ Use when

σ
σ known x̄ ± zα/2 √ population SD is given
n
s
σ unknown x̄ ± tα/2,n−1 √ only sample SD is given
n

Worked pattern

Given x̄ = 50, σ = 10, n = 25. A 95% CI is

 
10
50 ± 1.96 √ = 50 ± 3.92 = (46.08, 53.92).
25
Meaning: We are 95% confident that the true population mean lies in this interval.

3
Applied Statistics Concise Exam Notes

5. Hypothesis Testing
A hypothesis test asks whether sample evidence is strong enough to reject a claim.

Symbol Meaning
H0 null hypothesis: no change, no difference, claimed value
H1 alternative hypothesis: what the question wants evidence for
α significance level, usually 0.05
Test statistic standardized distance between sample result and H0

Five-step exam method

1. Write H0 and H1 .
2. Choose α.
3. Select the correct test statistic.
4. Compare with critical value or use p-value.
5. Conclude in words: reject or do not reject H0 .

5.1. Tests for one mean

Test Formula Use when

x̄ − µ0
One-sample Z Z= √ one mean, σ known
σ/ n
x̄ − µ0
One-sample t t= √ one mean, σ unknown
s/ n

Why this formula? The numerator is the difference from the claimed mean; the denominator is the
normal random error expected from sampling.

Worked pattern

Claim: µ = 1200. Given x̄ = 1250, σ = 100, n = 25.

1250 − 1200 50
Z= √ = = 2.5.
100/ 25 20

At 5% two-tailed level, critical value is ±1.96. Since 2.5 > 1.96, reject H0 .

4
Applied Statistics Concise Exam Notes

5.2. Tests for two means

Two-sample tests

If population SDs are known or samples are large:

( x̄1 − x̄2 ) − 0
Z= q .
σ12 /n1 + σ22 /n2

If variances are unknown but assumed equal:

(n1 − 1)s21 + (n2 − 1)s22 x̄1 − x̄2

S2p = , t= √ .
n1 + n2 − 2 Sp 1/n1 + 1/n2

If variances are not assumed equal:

x̄1 − x̄2
t= q .
s21 /n1 + s22 /n2

Core idea
Two-sample tests compare the observed difference x̄1 − x̄2 with the random error expected from
two independent samples. Independent variances add, so the denominator contains a sum.

6. Chi-square Tests
6.1. Goodness-of-fit test
Use when checking whether observed counts follow a theoretical distribution such as Poisson or bino-
mial.

Goodness-of-fit

(Oi − Ei )2
χ2 = ∑ Ei , d f = k − 1 − m.

Oi = observed count, Ei = expected count, k = number of final categories, m = number of parameters

estimated from data.

How to apply

1. Find expected probabilities from the proposed distribution.

2. Convert probabilities to expected counts: Ei = N pi .
3. Combine classes if any expected count is too small, usually below 5.
4. Compute χ2 and compare with table value.

Why use this formula? It measures total squared difference between observed and expected counts,
scaled by expected size.

6.2. Test of independence / association

Use when the data are in a contingency table.

5
Applied Statistics Concise Exam Notes

Independence test

(row total)(column total) (Oij − Eij )2

Eij =
grand total
, χ2 = ∑ Eij
, d f = (r − 1)(c − 1).

Conclusion wording

If calculated χ2 is larger than critical χ2 , reject H0 . Conclude: “there is evidence of association

between the two categorical variables.”

6.3. Chi-square test for one variance

Use when testing one population variance or standard deviation.

One variance test

( n − 1) s2
χ2 = , d f = n − 1.
σ02

Important exam note

If the question gives standard deviation, square it first. For example, s = 5 gives s2 = 25 and σ0 = 4
gives σ02 = 16.

7. F-Test for Two Variances

Use when comparing variation in two independent populations.

Two variance test

s21
H0 : σ12 = σ22 , F = , d f 1 = n1 − 1, quadd f 2 = n2 − 1.
s22
Usually place the larger sample variance on top so F ≥ 1.

Why F?

The ratio of two independent sample variances follows an F-distribution when the populations are
normal.

8. Correlation and Regression

8.1. Simple linear regression
Regression predicts a response variable Y from an explanatory variable X.

Least-squares regression line

Sxy
ŷ = a + bx, b= , a = ȳ − b x̄,
Sxx
( ∑ x )2 (∑ x )(∑ y)
Sxx = ∑ x2 − n
, Sxy = ∑ xy − n
.

Why these formulas? The line is chosen by least squares: it minimizes total squared prediction error
∑(y − ŷ)2 .

6
Applied Statistics Concise Exam Notes

Worked pattern

If b = 1.5 and a = 0.67, then

ŷ = 0.67 + 1.5x.
When x = 4,
ŷ = 0.67 + 1.5(4) = 6.67.
The slope means each one-unit increase in X increases predicted Y by 1.5 units.

8.2. Correlation
Correlation measures strength and direction of a linear relation.

−1 ≤ r ≤ 1.

r value Meaning
r>0 positive linear relationship
r<0 negative linear relationship
r≈0 weak or no linear relationship
|r | ≈ 1 strong linear relationship

Warning

Correlation does not prove causation. It only describes linear association.

9. Regression ANOVA
Regression ANOVA tests whether the regression model explains a significant amount of variation in Y.

Variation decomposition

SST = SSR + SSE,

SST = ∑(y − ȳ)2 , SSR = ∑(ŷ − ȳ)2 , SSE = ∑(y − ŷ)2 .

Source SS df MS Test
Regression SSR 1 MSR = SSR/1 F = MSR/MSE
Error SSE n−2 MSE = SSE/(n − 2)
Total SST n−1

Decision

H0 : β 1 = 0 versus H1 : β 1 ̸= 0.
If Fcalc > Fcritical , reject H0 . The regression is significant.

7
Applied Statistics Concise Exam Notes

10. Final Exam Decision Chart

Question clue What it means Use

“out of n”, success/failure fixed trials Binomial
“per hour/month/metre” count in interval Poisson
Measurement with mean and SD continuous bell-shape Normal √
Sample average X̄ mean of many observations Z = ( X̄ − µ)/(σ/ n)
Mean, σ known one population mean Z-test / Z-CI
Mean, σ unknown one population mean t-test / t-CI
Compare two means two groups two-sample Z or t
Observed vs expected counts fit to distribution Chi-square GOF
Two categorical variables association Chi-square independence
One variance/SD variation claim Chi-square variance test
Two variances/SDs compare spreads F-test
Predict Y from X relationship model Regression

Final takeaway

Most marks come from choosing the correct method, writing hypotheses clearly, substituting val-
ues into the right formula, and giving a conclusion in words. Always identify: parameter, sample
statistic, distribution/test, degrees of freedom and decision rule.