# Statistical Inference — Quick Reference Guide
## 1. Sampling and Sampling Distributions
- **Population vs Sample**: population = entire
group; sample = subset used to infer population
parameters.
- **Sampling methods**: simple random,
stratified, systematic, cluster sampling.
- **Sampling distribution**: distribution of a
statistic (e.g., sample mean) over repeated
samples.
- **Central Limit Theorem (CLT)**: for large n,
the sampling distribution of the sample mean is
approximately Normal(μ, σ²/n), regardless of
population shape.
- **Standard Error (SE)**: SE(x̄) = σ/√n (use s/
√n when σ unknown).
## 2. Point Estimation
- **Estimator vs Estimate**: estimator =
formula/rule (e.g., x̄); estimate = numeric
value from data.
- **Properties of good estimators**:
- *Unbiasedness*: E(θ̂) = θ
- *Consistency*: θ̂ → θ as n → ∞
- *Efficiency*: smallest variance among
unbiased estimators
- *Sufficiency*: uses all information in
sample about θ
- **Common estimators**: x̄ for μ, s² for σ², p̂
for population proportion.
- **Methods**: Method of Moments, Maximum
Likelihood Estimation (MLE).
## 3. Confidence Intervals (Interval
Estimation)
- **General form**: estimate ± (critical value)
× (standard error)
- **CI for mean (σ known)**: x̄ ± Z(α/2) · σ/√n
- **CI for mean (σ unknown)**: x̄ ± t(α/2, n−1)
· s/√n
- **CI for proportion**: p̂ ± Z(α/2) ·
√[p̂(1−p̂)/n]
- **CI for difference of means**: (x̄₁−x̄₂) ±
critical value · SE(difference)
- **Interpretation**: "We are 95% confident the
true parameter lies in this interval" (not a
probability statement about the fixed
parameter).
- Width of CI ↓ as n ↑, and ↑ as confidence
level ↑.
## 4. Hypothesis Testing – Fundamentals
- **Null (H₀) vs Alternative (H₁/Hₐ)**
hypotheses.
- **Type I error (α)**: rejecting true H₀.
**Type II error (β)**: failing to reject false
H₀.
- **Power** = 1 − β (probability of correctly
rejecting false H₀).
- **Test statistic, p-value, critical region**.
- **Decision rule**: reject H₀ if p-value < α,
or if test statistic falls in critical region.
- **One-tailed vs two-tailed tests**.
- **Steps**: state hypotheses → choose α →
compute test statistic → find p-value/critical
value → conclude.
## 5. Z & T Distributions
- **Z-distribution**: Standard Normal (mean 0,
sd 1); used when σ known or n large (n ≥ 30).
- **T-distribution**: used when σ unknown and n
small; has heavier tails than Z, shape depends
on degrees of freedom (df = n−1).
- As df → ∞, t-distribution → Z-distribution.
- **When to use which**:
- σ known → Z
- σ unknown, n large → Z (approx, using s)
- σ unknown, n small → t
## 6. One-Sample Tests for Mean
- **Hypotheses**: H₀: μ = μ₀ vs H₁: μ ≠ μ₀ (or
>, <)
- **Test statistic (σ known)**: Z = (x̄ −
μ₀)/(σ/√n)
- **Test statistic (σ unknown)**: t = (x̄ −
μ₀)/(s/√n), df = n−1
- Compare to critical value or compute p-value;
reject H₀ accordingly.
## 7. Two-Sample Tests for Means
- **Independent samples (equal variances)**:
pooled t-test
- t = (x̄₁−x̄₂)/√[Sp²(1/n₁+1/n₂)], df = n₁+n₂−2
- **Independent samples (unequal variances)**:
Welch's t-test (approx df formula)
- **Paired samples**: test on differences d =
x₁−x₂
- t = d̄/(s_d/√n), df = n−1
- Choose paired vs independent based on study
design (same subjects measured twice = paired).
## 8. Tests for Proportions
- **One-sample proportion test**: Z = (p̂ − p₀)/
√[p₀(1−p₀)/n]
- **Two-sample proportion test**: Z = (p̂₁−p̂₂)/
√[p̄(1−p̄)(1/n₁+1/n₂)], where p̄ is pooled
proportion.
- Requires np ≥ 5 and n(1−p) ≥ 5 for Normal
approximation to hold.
## 9. Chi-Square Tests
- **Goodness of Fit**: tests if observed
frequencies match an expected distribution.
- χ² = Σ (Oᵢ − Eᵢ)²/Eᵢ, df = k−1
- **Test of Independence**: tests association
between two categorical variables (contingency
table).
- df = (rows−1)(columns−1)
- **Test of Homogeneity**: compares
distribution of a categorical variable across
multiple populations.
- Assumptions: expected cell counts typically ≥
5.
## 10. ANOVA and Variance-Related Tests
- **One-Way ANOVA**: compares means across 3+
groups.
- H₀: μ₁ = μ₂ = ... = μₖ
- F = MSB/MSW (Mean Square Between / Mean
Square Within)
- df: between = k−1, within = N−k
- **Assumptions**: independence, normality,
homogeneity of variances (homoscedasticity).
- **Post-hoc tests** (if H₀ rejected): Tukey's
HSD, Bonferroni — to find which groups differ.
- **Two-Way ANOVA**: tests effects of two
factors and their interaction.
- **Variance tests**:
- **F-test**: compares two variances (σ₁² vs
σ₂²), F = s₁²/s₂²
- **Levene's/Bartlett's test**: tests
equality of variances across multiple groups
(used to check ANOVA assumptions).
---
### Quick Decision Map
| Scenario | Test |
|---|---|
| 1 mean, σ known | Z-test |
| 1 mean, σ unknown | t-test |
| 2 independent means | Independent t-test |
| 2 paired means | Paired t-test |
| 3+ means | ANOVA (F-test) |
| 1 proportion | Z-test for proportion |
| 2 proportions | Z-test (pooled) |
| Categorical association | Chi-square test |
| Compare variances | F-test / Levene's |