8.2: Inference for a Single Population Mean Using the Student's t-Distribution

Estimating Population Means When Population Standard Deviation Is Unknown

  • In prior procedures (such as Section 8.1), confidence intervals for population means were calculated from sample data under the prerequisite that the population standard deviation (σ\sigma) was known beforehand.

  • When σ\sigma is unknown, the population standard deviation is estimated using the sample standard deviation (ss), the calculation of which was established in prior work (Section 2.7).

  • Estimating σ\sigma with ss introduces significant uncertainty if the sample size is small, because the sample standard deviation may deviate substantially from the true population standard deviation.

  • Because of this potential discrepancy, standard normal methods cannot be used directly; a modified procedure substituting ss for σ\sigma is required.

Properties and Characteristics of the Student's t-Distribution

  • The Student's t-distribution is a family or collection of continuous probability distributions that approximate the standard normal distribution.

  • Compared to the standard normal distribution, the t-distribution possesses heavier or wider "tails," accounting for the additional variability introduced by estimating σ\sigma with ss.

  • Each specific t-distribution within the family is determined by a parameter called the degrees of freedom (df\text{df}).

  • Degrees of freedom quantify how closely a specific t-distribution matches the standard normal distribution:

    • Higher degrees of freedom yield distributions that more closely mirror the standard normal distribution (for example, a distribution with df=50\text{df} = 50 resembles the standard normal far more closely than one with df=5\text{df} = 5).

  • The Student's t-distribution is applicable whenever the underlying population is assumed to be normally distributed.

  • In inferential calculations, the degrees of freedom are defined by the sample size (nn):   df=n−1\text{df} = n - 1

  • An associated critical value (t-score, denoted tα/2t_{\alpha/2}) is located using a t-distribution table matching the calculated degrees of freedom (n−1n - 1) and the designated level of confidence.

  • The resulting t-score indicates how many estimated standard deviations away from the sample mean must be extended to capture the population mean within the specified confidence interval.

  • Standard reference tables for t-distribution critical values include:

    • https://www.sjsu.edu/faculty/gerstman/StatPrimer/t-table.pdf

    • https://www.stat.purdue.edu/~lfindsen/stat503/t-Dist.pdf

Mathematical Formulation of the Confidence Interval

  • The sample mean (denoted Xˉ\bar{X}) is approximately normally distributed:   Xˉ∼N(μ,σn)\bar{X} \sim N\left(\mu, \frac{\sigma}{\sqrt{n}}\right)

  • In confidence interval construction, the t-score functions analogously to the z-score used when σ\sigma is known.

  • The confidence interval for a single population mean (μ\mu) is expressed by the compound inequality:   xˉ−tα/2⋅sn<μ<xˉ+tα/2⋅sn\bar{x} - t_{\alpha/2} \cdot \frac{s}{\sqrt{n}} < \mu < \bar{x} + t_{\alpha/2} \cdot \frac{s}{\sqrt{n}}

    • xˉ\bar{x} represents the sample mean.

    • ss represents the sample standard deviation.

    • nn represents the sample size.

    • tα/2t_{\alpha/2} represents the critical t-score corresponding to df=n−1\text{df} = n - 1 and confidence level 1−α1 - \alpha.

    • μ\mu represents the unknown true population mean.

Procedural Steps and Case Examples

  • Golden Retriever Heights Example:

    • Objective: Construct a 95%95\% confidence interval for the average height of golden retrievers.

    • Sample Data (n=5n = 5):

    • 22.25′′22.25''

    • 23.5′′23.5''

    • 21′′21''

    • 22.5′′22.5''

    • 23.75′′23.75''

    • Step 1: Calculate the sample mean (xˉ\bar{x}) and the sample standard deviation (ss) from the five height measurements.

    • Step 2: Determine degrees of freedom (df=5−1=4\text{df} = 5 - 1 = 4) and find the critical value tα/2t_{\alpha/2} corresponding to a 95%95\% confidence level at df=4\text{df} = 4.

    • Step 3: Substitute xˉ\bar{x}, ss, n=5n = 5, and the critical t-score into the confidence interval inequality:     xˉ−tα/2⋅sn<μ<xˉ+tα/2⋅sn\bar{x} - t_{\alpha/2} \cdot \frac{s}{\sqrt{n}} < \mu < \bar{x} + t_{\alpha/2} \cdot \frac{s}{\sqrt{n}}

  • Emperor Penguin Weights Example:

    • Population Assumption: Emperor penguin weights are assumed to be normally distributed.

    • Sample Size: n=10n = 10 penguins.

    • Sample Mean (xˉ\bar{x}): 50.4 pounds50.4\,\text{pounds}.

    • Sample Standard Deviation (ss): 1 pound1\,\text{pound}.

    • Objective: Determine a 95%95\% confidence interval for the population weight (μ\mu) using df=10−1=9\text{df} = 10 - 1 = 9.

  • Apple Orchard Production Example:

    • Population Assumption: A tree's apple production is assumed to be normally distributed.

    • Sample Size: n=30n = 30 trees investigated.

    • Sample Mean (xˉ\bar{x}): 15.2 bushels15.2\,\text{bushels} per tree.

    • Sample Standard Deviation (ss): 0.5 bushels0.5\,\text{bushels}.

    • Objective: Determine a 98%98\% confidence interval for the population production (μ\mu) using df=30−1=29\text{df} = 30 - 1 = 29.