HL: advanced modelling/inference - Mathematics: Applications & Interpretation IB Study Notes

Advanced Statistical Modelling and Inference forms the pinnacle of IB Mathematics: Applications & Interpretation HL, combining probability distributions with hypothesis testing and confidence intervals.

Key Definitions:

Hypothesis Testing involves making decisions about population parameters based on sample data. The null hypothesis (H₀) represents the status quo or no effect, while the alternative hypothesis (H₁) represents what we're testing for. The significance level (α) is the probability of rejecting H₀ when it's true (Type I error), typically set at 0.05 or 0.01.

Confidence Intervals provide a range of plausible values for a population parameter. A 95% confidence interval means that if we repeated sampling infinitely, 95% of intervals would contain the true parameter.

Essential Formulas:

For population mean (σ known): CI = x̄ ± z* × (σ/√n)

For population mean (σ unknown): CI = x̄ ± t* × (s/√n)

For population proportion: CI = p̂ ± z* × √(p̂(1-p̂)/n)

Test Statistics:

• Z-test: z = (x̄ - μ₀)/(σ/√n) when σ is known
• t-test: t = (x̄ - μ₀)/(s/√n) when σ is unknown
• χ² goodness-of-fit: χ² = Σ[(O - E)²/E]

p-value represents the probability of obtaining results at least as extreme as observed, assuming H₀ is true. If p-value < α, reject H₀.

Memory Aid: "PHANTOM" - Parameter, Hypothesis, Alpha, Null distribution, Test statistic, Obtain p-value, Make decision.

Type II Error (β) occurs when we fail to reject a false null hypothesis. Power = 1 - β measures the test's ability to detect true effects.

Real-World Context: Quality Control in Manufacturing

Imagine a pharmaceutical company producing tablets that must contain exactly 500mg of active ingredient. Statistical inference helps determine whether production meets standards without testing every tablet.

Hypothesis Testing as Decision-Making: Think of hypothesis testing like a courtroom trial. The null hypothesis (H₀: μ = 500mg) is "innocent until proven guilty." We collect evidence (sample data) and decide whether evidence is strong enough to convict (reject H₀). The significance level α represents how certain we need to be—like requiring "beyond reasonable doubt."

Confidence Intervals in Climate Science: When researchers estimate global temperature increases, they report ranges like "1.5°C to 2.3°C increase with 95% confidence." This acknowledges uncertainty while providing actionable information for policy decisions.

The χ² Test in Genetics: Biologists use chi-squared tests to verify Mendelian inheritance ratios. If crossing two heterozygous plants should yield a 3:1 ratio, but observed data shows 142:38, does this support or refute the model?

Type I vs Type II Errors—Medical Analogy:

• Type I Error: Diagnosing a healthy patient as sick (false positive). Like a fire alarm going off when there's no fire—inconvenient but safe.
• Type II Error: Missing disease in a sick patient (false negative). Like a fire alarm failing during an actual fire—potentially catastrophic.

Power Analysis in Research: Before conducting expensive studies, researchers calculate required sample sizes. A clinical trial needs sufficient power (typically 80%) to detect meaningful differences. Underpowered studies waste resources; overpowered studies are unnecessarily expensive.

Analogy: Confidence intervals are like fishing nets—wider nets (lower confidence) catch more fish but include more debris; narrower nets (higher confidence) are more selective but might miss the target.