liveBook · Manning

Null hypothesis

As explained in Timeless Algorithms, The Foundational Ideas

A null hypothesis is a baseline statistical claim, commonly asserting that an effect or difference is absent.

A production process may appear to improve simply because of sampling variation. A null hypothesis, written H₀, supplies a baseline claim against which such evidence can be assessed. It commonly represents the status quo or the absence of an effect; the alternative hypothesis H₁ states the competing claim. section 4.1.1 section 10.2.3

The hypotheses are stated before examining the sample. Their form determines what the test can conclude and whether departures in one or both directions matter. A simple null hypothesis specifies a single parameter value; an alternative may be composite, covering a range of values. section 4.2.1 section 4.2.3

Checking beverage fill volumes

For containers advertised as holding 12 ounces, let μ be the population mean fill volume. The hypotheses H₀: μ = 12 and H₁: μ ≠ 12 compare the target with deviations in either direction, calling for a two-tailed test. Evidence sufficient to reject H₀ may prompt investigation of the production process. section 4.2.1

A p-value is the probability, assuming H₀ is true, of obtaining a result at least as extreme as the observed result. It is compared with a predetermined significance level α: a p-value at or below the threshold leads to rejection of H₀; otherwise, the decision is to fail to reject it. The p-value is not the probability that H₀ is true. section 10.2.3 section 4.1.6

Neither decision is proof. Rejecting a true H₀ is a Type I error; failing to reject a false H₀ is a Type II error. The significance level controls the long-run probability of a Type I error, while statistical power influences the probability of a Type II error. Choosing α requires judgment about the consequences of these errors. section 4.1.6 section 4.1.2

A clinical trial near the threshold

For a drug compared with a historical recovery rate of 50%, H₀: p = 0.50 and H₁: p > 0.50 express a one-tailed test for improvement. If the evidence falls just short of the predetermined significance threshold, failing to reject H₀ does not prove that the recovery rate remains 50%. section 4.2.3

How Timeless Algorithms, The Foundational Ideas covers it

How deep it goes
Develops hypothesis specification, significance thresholds, decisions, and error risks through worked examples.
Examples
Illustrates the null with beverage fill volumes, a clinical trial recovery rate, and a manufacturing defect rate.
What you can do after
Evaluate and understand test decisions without treating rejection or failure to reject as proof.
How it connects
Connects hypothesis testing to alternative hypotheses, p-values, significance levels, Type I and Type II errors, and statistical power.

Also explained in

Compare how 6 books explain null hypothesis →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Timeless Algorithms, The Foundational Ideas. The section links point to the parts of the book it draws on.