Sampling Distributions of the Sample Mean and Foundations of Statistical Inference

Course Assessment and Curricular Architecture

  • The first examination is substantially easier than the second examination; while some students find the second exam more intuitive due to its procedural structure, it requires distinguishing among distinct problem types rather than addressing straightforward, single-chapter prompts.

  • All questions on the second examination follow a consistent procedural methodology, with the present topic serving as the final foundational building block necessary to execute formal statistical analyses.

  • Scoring records reflect an exam evaluated out of 8080 points, supplemented by a 77-point score adjustment that has been integrated into grade records.

  • The Chapter 15 reading quiz is due tonight; the previously considered 55-point bonus opportunity ended after Chapter 3 following an unfulfilled conditional opportunity.

  • The instructional material serves as a direct extension of Chapter 3 (normal distributions) and the continuous probability distributions covered in Chapter 12, providing the analytical bridge to formal statistical inference.

  • Standardized testing policy across higher education is experiencing an institutional shift, with numerous undergraduate institutions reinstating the SAT and ACT, and graduate programs consistently requiring the GRE and MCAT.

Principles of Statistical Inference: Parameters and Statistics

  • Statistical inference is the process of drawing conclusions about an unknown population based on information obtained from an observed sample.

  • Numerical descriptors follow a strict conceptual distinction based on their origin:

    • A parameter is a numerical value that describes an entire population (mnemonic: Population to Parameter).
    • A statistic is a numerical value computed from a subset of the population, known as a sample (mnemonic: Sample to Statistic).
  • The fundamental challenge of statistics arises because calculating true population parameters directly is generally impossible or impractical due to resource, population size, or accessibility constraints; researchers instead collect sample data, compute sample statistics, and infer back to the population parameter.

  • Primary point estimators currently evaluated:

    • The population mean is denoted by the Greek letter μ\mu.
    • The sample mean is denoted by xˉ\bar{x} (X-bar), which serves as the empirical estimate of μ\mu.

Foundational Assumptions: Representativeness and the Law of Large Numbers

  • Valid statistical inference relies on two indispensable core principles:

  • Principle 1: Sample Representativeness:

    • A sample must accurately reflect the characteristics of the target population to permit valid generalizations.
    • Example: An investigation designed to make inferences regarding oak trees cannot derive its data exclusively from aspen trees, as differing species characteristics invalidate the inference.
    • Contextual generalizability: Certain analytical contexts permit minor sampling scope extensions, such as utilizing a specific institutional cohort of college students to make broader generalizations regarding the general college student body.
    • The Simple Random Sample (abbreviated as SRS) provides the operational standard; whenever an SRS is acquired, the sample is assumed to be representative of the target population, justifying the use of xˉ\bar{x} as an unbiased estimator for μ\mu.
  • Principle 2: The Law of Large Numbers:

    • The Law of Large Numbers states that as the number of observations drawn from a population increases (as sample size nn increases), the sample statistic xˉ\bar{x} converges closer and closer to the true population parameter μ\mu.
    • Classroom scenario: In an academic course with a total roster of 2424 enrolled students, calculating a class average based on a subset of 55 students yields an estimate closer to the true class average than an average derived from 33 students; a sample of 1010 students yields a closer estimate than 55 students; and a sample of 1515 students yields a closer estimate than 1010 students.
    • Core statistical consequence: Increased sample size systematically reduces estimation error and draws sample estimators closer to the objective parameter value.

The Sampling Distribution of the Sample Mean (xˉ\bar{x})

  • The analytical focus transitions from evaluating single, isolated observations (xx) to evaluating the arithmetic mean (xˉ\bar{x}) of a collection of nn observations.

  • Definition: The sampling distribution of xˉ\bar{x} represents the theoretical distribution produced if every possible unique sample of a fixed size nn were repeatedly drawn from a population of size NN (e.g., drawing all possible unique samples of size n=20n = 20 from a population of N=100N = 100), computing the respective xˉ\bar{x} for each sample, and plotting every resulting xˉ\bar{x} into a frequency histogram.

  • Mathematical properties under a normal parent population:

    • If an individual observation xx follows a normal distribution with mean μ\mu and standard deviation σ\sigma:

x∼N(μ,σ)x \sim N(\mu, \sigma)

  • Then the sample mean xˉ\bar{x} derived from an SRS of size nn is also normally distributed with the exact same mean μ\mu, but with an adjusted standard deviation denoted as σxˉ\sigma_{\bar{x}}:

xˉ∼N(μ,σn)\bar{x} \sim N\left(\mu, \frac{\sigma}{\sqrt{n}}\right)

  • Dispersion dynamics and variance adjustment:
    • The mean of the sampling distribution remains identical to the population mean (μxˉ=μ\mu_{\bar{x}} = \mu).
    • The standard deviation of the sample mean (the standard error) is reduced by dividing the population standard deviation by the square root of the sample size:

σxˉ=σn\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}}

  • Justification: Standard deviation measures data variability. A single individual randomly selected from a group exhibits substantial variability relative to the whole, whereas the pooled average of 1010 individuals smooths out individual extremes and is systematically closer to the true population average.
  • Sample means are strictly less variable than individual observations. In mathematical terms, dividing a constant numerator σ\sigma by an increasingly large denominator n\sqrt{n} forces the standard error to shrink toward zero as nn expands.

The Central Limit Theorem (CLT)

  • The Central Limit Theorem establishes that when the sample size nn is sufficiently large, the sampling distribution of the sample mean xˉ\bar{x} will be approximately normal with mean μ\mu and standard deviation σn\frac{\sigma}{\sqrt{n}}, regardless of the underlying distribution of the original population data.

  • Core theorem implications:

    • The analyst does not need the underlying individual measurements to be normally distributed, provided the sample size is adequately large.
    • The distribution of sample averages possesses two defining structural characteristics relative to single observations: sample means are consistently less variable and significantly more normal in distribution shape than individual observations.

Standard Normal Calculations: Unifying Individual Observations and Sample Means

  • Standard normal conversion formula for a single observation (n=1n = 1):

z=x−μσz = \frac{x - \mu}{\sigma}

  • Standard normal conversion formula for a sample mean (n>1n > 1):

z=xˉ−μσnz = \frac{\bar{x} - \mu}{\frac{\sigma}{\sqrt{n}}}

  • Unified structural identity:
    • The single-observation formula is simply a special case of the sample mean formula.
    • Setting the sample size to n=1n = 1 yields:

xˉ=x\bar{x} = x

n=1=1\sqrt{n} = \sqrt{1} = 1

σn=σ1=σ\frac{\sigma}{\sqrt{n}} = \frac{\sigma}{1} = \sigma

  • Substituting these values yields the single-observation equation directly. Consequently, the sample mean formula functions as a universal standardized calculation for any sample size n≥1n \ge 1.

    • Computational calculator entry protocol:
  • Numerator expressions must be enclosed in parentheses: (x̄ - μ).

  • Denominator expressions must be enclosed in independent parentheses: (σ / √(n)).

  • Combined calculator input syntax: (x̄ - μ) / (σ / √(n)).

  • Square root generation on standard graphing calculators requires pressing the 2nd key followed by the x² key.

Standard Normal Table (Table A) Procedures and Boundary Conditions

  • Cumulative probability look-up conventions:
    • All computed zz-scores must be rounded to two decimal places prior to referencing Table A.
    • Less than probabilities (P(Z<z)P(Z < z)): Read the cumulative proportion directly from the intersection of the primary zz-row and the second decimal column.
    • Greater than probabilities (P(Z>z)P(Z > z)): Subtract the cumulative Table A table entry from 11:

P(Z>z)=1−P(Z<z)P(Z > z) = 1 - P(Z < z)

  • Between probabilities (P(z1<Z<z2)P(z_1 < Z < z_2)): Look up the cumulative probability for both the upper score z2z_2 and lower score z1z_1, then subtract the lower cumulative probability from the higher cumulative probability:

P(z1<Z<z2)=P(Z<z2)−P(Z<z1)P(z_1 < Z < z_2) = P(Z < z_2) - P(Z < z_1)

  • Any calculated probability must strictly result in a value between 00 and 11, never negative and never exceeding 11.

    • Boundary conditions for extreme zz-scores:
  • Dividing the standard deviation by n\sqrt{n} yields progressively smaller standard errors as nn increases, frequently generating extreme test statistics and zz-scores that fall outside the printed boundaries of Table A.

  • Upper extreme threshold: If a calculated zz-score exceeds 3.493.49 (z>3.49z > 3.49), the cumulative proportion reaches the asymptotic ceiling of the distribution; record the cumulative probability directly as 11 (or 1.00001.0000).

  • Lower extreme threshold: If a calculated zz-score falls below −3.49-3.49 (z<−3.49z < -3.49), the cumulative proportion reaches the lower asymptote; record the cumulative probability directly as 00 (or 0.00000.0000).

Comprehensive Case Studies and Step-by-Step Applications

  • Case Study 1: Medical College Admission Test (MCAT) Distribution Analysis:

    • MCAT scores follow a normal distribution with population mean μ=500\mu = 500 and population standard deviation σ=10.4\sigma = 10.4.

    • Part (a): Probability that a single randomly selected student scores less than 506506

    • Parameter identification: Individual observation (n=1n = 1), data value x=506x = 506.

    • Formula application:

z=506−50010.4=610.4≈0.58z = \frac{506 - 500}{10.4} = \frac{6}{10.4} \approx 0.58

  • Calculator syntax: (506 - 500) / 10.4.
  • Table A look-up: Locate row 0.5 and column 0.08, which yields a cumulative proportion of 0.71900.7190.
  • Probability statement:

P(X<506)=0.7190P(X < 506) = 0.7190

  • Substantive interpretation: There is approximately a 71.90%71.90\% (or roughly 72%72\%) probability that a randomly selected applicant scores below 506506.

    • Part (b): Probability that a random sample of 2525 students achieves a mean score greater than 498498
  • Parameter identification: Sample size n=25n = 25, sample mean target xˉ=498\bar{x} = 498.

  • Standard error determination:

σxˉ=10.425=10.45=2.08\sigma_{\bar{x}} = \frac{10.4}{\sqrt{25}} = \frac{10.4}{5} = 2.08

  • Standardized score calculation:

z=498−50010.425=−22.08≈−0.96z = \frac{498 - 500}{\frac{10.4}{\sqrt{25}}} = \frac{-2}{2.08} \approx -0.96

  • Calculator syntax: (498 - 500) / (10.4 / √(25)).
  • Table A look-up: Locate row -0.9 and column 0.06, which provides a cumulative area of 0.16850.1685.
  • Greater than transformation:

P(Xˉ>498)=1−P(Z<−0.96)=1−0.1685=0.8315P(\bar{X} > 498) = 1 - P(Z < -0.96) = 1 - 0.1685 = 0.8315

  • Substantive interpretation: There is an 83.15%83.15\% (or approximately 83%83\%) probability that the mean score of a sample of 2525 students will exceed 498498.

    • Part (c): Probability that a random sample of 2525 students achieves a mean score between 495495 and 505505
  • Parameter identification: Sample size n=25n = 25, lower threshold xˉ1=495\bar{x}_1 = 495, upper threshold xˉ2=505\bar{x}_2 = 505.

  • Lower bound zz-score transformation:

z1=495−50010.425=−52.08≈−2.40z_1 = \frac{495 - 500}{\frac{10.4}{\sqrt{25}}} = \frac{-5}{2.08} \approx -2.40

  • Upper bound zz-score transformation:

z2=505−50010.425=52.08≈2.40z_2 = \frac{505 - 500}{\frac{10.4}{\sqrt{25}}} = \frac{5}{2.08} \approx 2.40

  • Table A look-up for upper bound: Row 2.4, column 0.00 yields 0.99180.9918.
  • Table A look-up for lower bound: Row -2.4, column 0.00 yields 0.00820.0082.
  • Between probability calculation:

P(495<Xˉ<505)=P(Z<2.40)−P(Z<−2.40)=0.9918−0.0082=0.9836P(495 < \bar{X} < 505) = P(Z < 2.40) - P(Z < -2.40) = 0.9918 - 0.0082 = 0.9836

- Substantive interpretation: There is a 98.36%98.36\% (or approximately 98%98\%) probability that the sample mean across 2525 students falls within the interval between 495495 and 505505, demonstrating the high condensation of mean values relative to single scores.
  • Case Study 2: Gestational Diabetes Post-Prandial Glucose Screening:

    • Glucose levels measured one hour after ingestion of a standardized sugary beverage follow a normal distribution with population mean μ=122 mg/dL\mu = 122\,\text{mg/dL} and population standard deviation σ=12 mg/dL\sigma = 12\,\text{mg/dL}.

    • Part (a): Probability of an individual glucose reading exceeding 130 mg/dL130\,\text{mg/dL}

    • Evaluates an individual reading (n=1n = 1), targeting x>130x > 130.

    • Standardized score calculation:

z=130−12212=812≈0.67z = \frac{130 - 122}{12} = \frac{8}{12} \approx 0.67

  • Standard right-tail procedure: Look up z=0.67z = 0.67 on Table A and compute the complement 1−P(Z<0.67)1 - P(Z < 0.67).

    • Part (b): Probability that the mean glucose level across measurements collected on 44 separate days is less than 120 mg/dL120\,\text{mg/dL}
  • Evaluates a sample mean with sample size n=4n = 4, standard error σn=124=122=6\frac{\sigma}{\sqrt{n}} = \frac{12}{\sqrt{4}} = \frac{12}{2} = 6.

  • Standardized score calculation:

z=120−122124=−26≈−0.33z = \frac{120 - 122}{\frac{12}{\sqrt{4}}} = \frac{-2}{6} \approx -0.33

  • Direct left-tail procedure: Look up z=−0.33z = -0.33 directly on Table A to obtain P(Xˉ<120)P(\bar{X} < 120).

    • Part (c): Probability that the mean glucose level across 44 separate days falls between 125 mg/dL125\,\text{mg/dL} and 135 mg/dL135\,\text{mg/dL}
  • Evaluates an interval between two sample means (n=4n = 4):

z1=125−122124=36=0.50z_1 = \frac{125 - 122}{\frac{12}{\sqrt{4}}} = \frac{3}{6} = 0.50

z2=135−122124=136≈2.17z_2 = \frac{135 - 122}{\frac{12}{\sqrt{4}}} = \frac{13}{6} \approx 2.17

  • Interval procedure: Determine the cumulative probability corresponding to the larger zz-score (z2=2.17z_2 = 2.17), determine the cumulative probability for the smaller zz-score (z1=0.50z_1 = 0.50), and subtract the smaller value from the larger value to preserve a strictly positive probability.

Questions & Discussion

  • Question: What causes an input error on the calculator when evaluating expressions containing roots in the denominator?

    • Answer: Input errors commonly occur when invoking the general index root function (yx\sqrt[x]{y}) rather than the standard square root function (x\sqrt{x}). On standard graphing calculators, the dedicated square root is activated by pressing the 2nd key followed by the x² button. When entering the denominator expression, wrap the terms in parentheses, entering 12 / √(4) or 10.4 / √(25).
  • Question: When finding the area between two distinct standardized scores (such as z=0.50z = 0.50 and z=2.17z = 2.17), which table value is looked up first?

    • Answer: Always find the cumulative proportion corresponding to the larger zz-score first, then subtract the cumulative proportion of the smaller zz-score from it. This guarantees that the subtraction operation yields a valid, non-negative probability value situated between 00 and 11.
  • Question: Is the 55-point extra credit bonus still accessible for the Chapter 15 reading quiz assignment?

    • Answer: The extra credit opportunity was rescinded after Chapter 2, renewed on a conditional basis, and officially terminated after Chapter 3 without reinstatement.