Hypothesis testing for the mean: known sigma

The video introduces hypothesis testing for the mean, specifically when the population standard deviation (sigma) is known. It differentiates hypothesis testing from confidence intervals, stating that hypothesis tests are useful for a single parameter value of interest.

An example is presented where a chocolate bar manufacturer claims an average weight of 350g, but a sample of 10 bars has a mean of 347g. Two possible explanations for this difference are discussed: the advertised mean is wrong, or the sample mean is low by random chance. The goal of hypothesis testing is to distinguish between these possibilities.

The central limit theorem is introduced as the basis for hypothesis testing for the population mean. It states that the sample mean (xbar) is approximately normal, and its z-score can be calculated using the formula: (xbar - mu) / (sigma / sqrt(n)), which follows a standard normal distribution.

The first step in a hypothesis test is to assume the claimed parameter value is true, which is called the null hypothesis (H0). In the example, H0: mu = 350g. The alternative hypothesis (HA) is what we suspect might be true, contradicting the null, e.g., HA: mu < 350g. Hypotheses should be set before data collection.

The test uses the sample mean to judge between the hypotheses. The probability of obtaining a sample mean as extreme or more extreme than observed, assuming the null hypothesis is true, is called the p-value. A low p-value suggests stronger evidence against the null hypothesis.

For the chocolate bar example, with a known population standard deviation (sigma) of 4g, a z-score (test statistic) of -2.37 is calculated. This z-score quantifies how far the sample mean is from the null hypothesis prediction. The p-value, calculated using 'P norm' in R, is approximately 0.0089.

A p-value of 0.0089 means that if the true mean weight is 350g, a sample mean of 347g or less would only occur 0.89% of the time. This low probability leads to two possibilities: something unusual happened by chance, or the null hypothesis is false. Given the very low p-value, the null hypothesis is rejected in favor of the alternative, concluding that the true mean weight is likely less than 350g.

The video concludes by posing further questions for discussion in subsequent videos: how to define a 'small enough' p-value (leading to statistical significance), the difference between one-sided and two-sided alternative hypotheses, and how to proceed when the population standard deviation (sigma) is unknown.