Summary of "Introduction to Hypothesis Testing|Statistics|BBA|BCA|B.COM|B.TECH|Dream Maths"

Overview

The video (Dream Maths, instructor Bharti) introduces hypothesis testing in statistics, explaining key ideas, terminology, and the standard step-by-step procedure used to solve hypothesis-testing problems. Simple real-life examples (e.g., salaries in South Indian companies; class size = 90) illustrate forming a hypothesis, deciding directionality, and determining acceptance/rejection regions.

Main ideas and concepts

Hypothesis (informal): an assumption or tentative statement about a population based on sample observations (e.g., “salaries in South Indian companies are high”).
Population vs. sample:
- Population: the entire group of interest.
- Sample: the subset observed. Hypothesis testing draws conclusions about a population from sample information.
Null hypothesis (H0): the hypothesis assumed true at the start of testing. Written in statement/mathematical form (e.g., H0: μ = μ0). Always state H0 first.
Alternative hypothesis (H1 or Ha): the statement opposite to H0. Types:
- One-tailed (directional): H1: parameter > value or H1: parameter < value (rejection region on one tail).
- Two-tailed (non-directional): H1: parameter ≠ value (rejection regions split across both tails).
Type I error (α): rejecting H0 when it is actually true. α denotes the probability of this error.
Type II error (β): failing to reject H0 when H0 is actually false. β denotes that probability.
Power of a test: 1 − β, the probability of correctly rejecting a false H0.
Level of significance (α): the chosen acceptable probability of making a Type I error. Common choices: 0.05 (default if not stated), 0.01, 0.10. For example, α = 0.05 corresponds to 95% confidence.
Critical region (rejection region) and acceptance region: ranges of the test statistic derived from the sampling distribution (commonly the standard normal / z). If the computed statistic falls in the rejection region, reject H0; otherwise fail to reject H0.

Step-by-step methodology for hypothesis testing

State the null hypothesis H0 (assume this is true initially).
State the alternative hypothesis H1 (the opposite of H0) and decide whether H1 is:
- One-tailed right (H1: parameter > value),
- One-tailed left (H1: parameter < value),
- Two-tailed (H1: parameter ≠ value).
Choose the level of significance α (given in the problem; if not, default α = 0.05).
Select the appropriate test statistic (e.g., z for large-sample mean tests or when population σ is known; t for small samples with σ unknown).
Calculate the test statistic from the sample data.
Determine the critical value(s) corresponding to α and the test type (one- or two-tailed). These define the rejection region(s).
Compare the computed statistic to the critical value(s):
- If the statistic lies in the rejection (critical) region → reject H0.
- If the statistic lies in the acceptance region → fail to reject H0.
State the conclusion in context (interpret in terms of the original problem).

Notes on tails and rejection regions

One-tailed test: alternative uses a single inequality (>, <). The rejection region is entirely in one tail (right tail for H1: >, left tail for H1: <).
Two-tailed test: alternative uses ≠. The rejection region is split into two equal tails (α/2 in each tail).
The direction (sign in H1) determines which tail(s) you use to find critical value(s).

Tip: Always decide the directionality (one- vs two-tailed) before looking at the sample statistic, because it determines the rejection region.

Common critical z-values (standard normal)

Two-tailed (critical value = z_{1−α/2}):

α = 0.10 → z ≈ 1.65
α = 0.05 → z ≈ 1.96
α = 0.01 → z ≈ 2.58
α = 0.001 → z ≈ 3.29

One-tailed (critical value = z_{1−α}):

α = 0.10 → z ≈ 1.28
α = 0.05 → z ≈ 1.64
α = 0.01 → z ≈ 2.33
α = 0.001 → z ≈ 3.09

These values come from the standard normal distribution and are used to set the rejection region boundaries.

Examples used in the lesson

Salaries in South Indian companies: forms a hypothesis from observing a few companies to show how a speculative assumption becomes a testable null hypothesis.
Class size = 90: illustrates one-tailed tests (H1: > 90 or H1: < 90) vs two-tailed tests (H1: ≠ 90) and how rejection regions fall on the right tail, left tail, or both.

Practical tips emphasized

If α is not specified in a problem, default to α = 0.05 (95% confidence).
Learn/memorize the common z critical values for quick decision-making.
The testing procedure is formulaic—once H0, H1, and α are set: compute the statistic → compare to critical region → conclude.
Interpret conclusions in the context of the problem (don’t just state “reject” or “accept” without context).