Statistical Inference and Sampling Distributions

Course Deadlines and Topic Roadmap

• Assignment Deadlines:

  • Connect Homework Assignment: Due on Wednesday of next week.
  • Short Report: Due on Wednesday of next week.

• Fundamental Review of Normal Distribution Calculations:

  • The normal distribution is a symmetric, bell-shaped curve defined by its mean and standard deviation.
  • In Excel, probabilities associated with continuous normal distributions are evaluated using the cumulative distribution function:     =NORM.DIST(x, mean, standard_dev, cumulative)
  • The x parameter specifies the value of interest, mean represents the population mean $\mu$, standard_dev represents standard deviation $\sigma$, and cumulative is set to TRUE (or 1) to compute left-tail cumulative probabilities $P(X \le x)$.

• Scope of Statistical Inference:

  • Statistical inference involves utilizing sample statistics to formulate statements, estimates, or probabilistic assertions about an unobserved population parameter.
  • Key inferential techniques build upon probability theory and include:
  • Confidence Intervals: Constructing a range of plausible values for a population parameter based on sample statistics.
  • Hypothesis Testing: Evaluating specific claims or assertions regarding population parameters using sample evidence.

• Prerequisite Foundation:

  • Conducting statistical inference requires establishing the theoretical framework of the sampling distribution of sample means.

Mathematical Notation for Populations and Samples

• Population Parameters versus Sample Statistics:

  • Population Mean: $\mu$
  • Population Standard Deviation: $\sigma$
  • Population Size (Total Number of Observations): $N$
  • Sample Mean (Sample Average): $\bar{x}$
  • Sample Standard Deviation: $s$
  • Sample Size (Number of Observations in Sample): $n$

• Conceptual Definitions:

  • Population: The comprehensive set containing all individual elements, observations, or subjects of interest.
  • Sample: A representative subgroup or subset selected from the larger population.
  • Statistical Inference: The formal process of extrapolating sample observations to draw conclusions regarding entire populations under uncertainty.

Theoretical Foundations of Sampling Distributions

• Concept of the Sampling Distribution of Sample Means:

  • Obtained by repeatedly drawing random samples of fixed size $n$ from a given population, calculating the sample mean $\bar{x}$ for each draw, and plotting the probability distribution of all possible sample means.
  • Focuses entirely on evaluating average characteristics rather than individual population values.

• Representative Population Example (Hourly Wages):

  • Consider a population consisting of $N = 7$ employees with hourly earnings distributed as follows:
  • $3$ employees earning $\$14$ per hour.
  • $3$ employees earning $\$16$ per hour.
  • $1$ employee earning $\$18$ per hour.
  • Population mean $\mu$ and population measures of dispersion can be derived directly from these $7$ observations.

• Constructing the Sampling Distribution for Sample Size $n = 2$:

  • Draw all possible sample combinations of size $n = 2$ employees from the population of $N = 7$
  • Compute the sample mean $\bar{x}$ for every pair.
  • Summarize the frequency distribution of sample averages:
  • $3$ distinct samples yield a sample mean of $\bar{x} = \$14$
  • $9$ distinct samples yield a sample mean of $\bar{x} = \$15$
  • Remaining sample combinations yield corresponding means up to $\bar{x} = \$18$
  • Plotting these sample averages as a frequency histogram produces the visual representation of the sampling distribution of sample means.

• Mathematical Properties of the Sampling Distribution:

  • Mean of Sample Means ($\mu_{\bar{x}}$): The average of all sample means across all possible samples of size $n$ is exactly equal to the population mean:     $\mu_{\bar{x}} = \mu$
  • Standard Error of the Mean ($\text{SE}$): The standard deviation of the sampling distribution measures the dispersion of sample averages around the true population mean. It is defined as:     $\text{SE} = \frac{\sigma}{\sqrt{n}}$

Precision, Comparative Static Analysis, and Sample Size Effects

• Comparative Static Analysis of Sample Size $n$:

  • Standard Error formula:     $\text{SE} = \frac{\sigma}{\sqrt{n}}$
  • Holding the population standard deviation $\sigma$ constant, increasing the sample size $n$ ($n \rightarrow \infty$) increases the denominator $\sqrt{n}$, thereby shrinking the overall standard error ($\text{SE} \rightarrow 0$).

• Mathematical Proof and Numerical Comparison:

  • Case 1: Pulling samples of size $n = 25$:     $\text{SE}_{25} = \frac{\sigma}{\sqrt{25}} = \frac{\sigma}{5}$
  • Case 2: Pulling samples of size $n = 100$:     $\text{SE}_{100} = \frac{\sigma}{\sqrt{100}} = \frac{\sigma}{10}$
  • Comparison:     $\frac{\sigma}{5} > \frac{\sigma}{10}$
  • Increasing sample size from $n = 25$ to $n = 100$ cuts the standard error in half.

• Practical Implications:

  • Greater sample sizes yield increased estimation precision.
  • As $n$ grows, the sampling distribution shrinks visually and concentrates more tightly around the central population mean $\mu$
  • This mathematical relationship explains the foundational emphasis on large sample sizes and modern data collection strategies in empirical research.

Central Limit Theorem (CLT)

• Core Theorem Definition:

  • The Central Limit Theorem dictates that for any population distribution with a finite mean $\mu$ and standard deviation $\sigma$, the sampling distribution of the sample mean $\bar{x}$ approaches a normal distribution with mean $\mu_{\bar{x}} = \mu$ and standard error $\text{SE} = \frac{\sigma}{\sqrt{n}}$ as sample size $n$ becomes large.

• Sample Size Threshold:

  • The universally accepted practical rule of thumb states that sample sizes of $n \ge 30$ are sufficient to guarantee that the sampling distribution of sample means is normally distributed, regardless of the original population shape.

• Conceptual Example:

  • Given a population of $N = 100$ total observations, repeatedly drawing subgroups of size $n = 30$, recording each sample mean $\bar{x}$, and plotting all resulting sample means yields a symmetric bell-shaped normal curve.

Computational Applications and Quality Control Examples

• Case Study Context: Quality Control of Coffee Package Filling:

  • Target packaging weight: $\mu = 500\,\text{g}$
  • Process standard deviation: $\sigma = 8\,\text{g}$
  • Inspection threshold: Testing for packages weighing less than $495\,\text{g}$

• Evaluation Type 1: Single Random Observation Probability (Individual Item Analysis)

  • Problem Statement: Calculate the probability that a single randomly selected bag weighs less than $495\,\text{g}$.
  • Relevant Parameters:
  • Mean: $\mu = 500\,\text{g}$
  • Standard Deviation: $\sigma = 8\,\text{g}$
  • Target Value: $x = 495\,\text{g}$
  • Probability Expression:     $P(X < 495)$
  • Excel Formula:     =NORM.DIST(495, 500, 8, TRUE)
  • Result:     $P(X < 495) \approx 0.266$ (or $26.6\%$

• Evaluation Type 2: Sample Average Probability (Sampling Distribution Analysis)

  • Problem Statement: Calculate the probability that the average weight $\bar{x}$ of a random sample of $n = 9$ bags is less than $495\,\text{g}$.
  • Relevant Parameters:
  • Mean of Sampling Distribution: $\mu_{\bar{x}} = 500\,\text{g}$
  • Sample Size: $n = 9$
  • Standard Error Calculation:       $\text{SE} = \frac{\sigma}{\sqrt{n}} = \frac{8}{\sqrt{9}} = \frac{8}{3} \approx 2.667\,\text{g}$
  • Target Value: $\bar{x} = 495\,\text{g}$
  • Probability Expression:     $P(\bar{x} < 495)$
  • Excel Formula:     =NORM.DIST(495, 500, 8/SQRT(9), TRUE)
  • Interpretation and Comparison:
  • The calculated probability $P(\bar{x} < 495)$ drops significantly below $26.6\%$ because the standard error of averages ($\text{SE} = \frac{8}{3}\,\text{g}$) is much smaller than individual standard deviation ($\sigma = 8\,\text{g}$).
  • If sample size were reduced to $n = 2$, standard error would be $\frac{8}{\sqrt{2}}\,\text{g}$, yielding a probability higher than $n = 9$ but lower than single bag selection ($n = 1$).
  • Higher sample sizes ($n = 36$) further compress standard error to $\frac{8}{\sqrt{36}} = \frac{8}{6} \approx 1.333\,\text{g}$, making extreme sample average deviations extremely rare.