2.2 )The Normal Population Model and Standardization

Four-Step Process of Statistical Analysis

  • Step 1: Identify Population and Parameter of Interest
    • Research Question Example: What is the average age at which BYU students find out Santa Claus isn't real? Specifically, is the average age at which BYU students find out Santa isn't real greater than 88?
    • Parameter: The mean age at which all BYU students find out Santa Claus isn't real, represented by the Greek letter μ\mu.
  • Step 2: Collect Data
    • Data is gathered from a target subset or sample of the population.
    • Sampling Example: A convenience sample consisting of 17631763 BYU students enrolled in a class who completed a student survey.
  • Step 3: Posit a Statistical Model Based on Sample Information
    • Exploration of sample data yields an understanding of the underlying population distribution shape.
  • Step 4: Draw Inference About the Population Using the Model
    • Statistical conclusions about population parameters (such as μ\mu) are drawn from the established theoretical model.

Statistical Models and Parameter Notation

  • Definition of Statistical Model: A mathematical representation describing the theoretical shape and characteristics of a population distribution.
  • Notation Conventions:
    • Greek Letters: Represent population-level values (parameters). Examples include μ\mu for population mean and σ\sigma for population standard deviation.
    • Roman Letters: Represent sample-level values (statistics) or individual dataset observations. Examples include ss for sample standard deviation and YY for individual data points.
  • Primary Purposes of a Statistical Model:
    • To describe the broader population distribution rather than merely summarizing the observed sample.
    • To enable formal statistical inferences about unknown population parameters based on sample statistics.
    • To calculate theoretical probabilities associated with population occurrences.

Mathematical Formulation of the Normal Population Model

  • Formal Notation:Y∼N(μ,σ)Y \sim \mathcal{N}(\mu, \sigma)
    • YY: An individual observation within the dataset.
    • ∼\sim: Symbol representing "comes from" or "is distributed as".
    • N\mathcal{N}: Denotes a population that follows a normal density distribution.
    • μ\mu: The (unknown) population mean.
    • σ\sigma: The (unknown) population standard deviation (sigma).

Normal population model notation breakdown

  • Conceptual Meaning of the Mathematics:
    • If every single student in the entire target population (e.g., all BYU students) were surveyed regarding the age they learned about Santa Claus, a density plot of those values would produce a smooth, bell-shaped normal curve.
    • The population-wide mean age is μ\mu, and the population-wide standard deviation of ages is σ\sigma.

Normal population density plot curve

Assessing Applicability of the Normal Population Model

  • Determining Population Model Appropriateness from Sample Data:
    • Because true population data is rarely fully observed, model selection relies on evaluating sample characteristics.
    • Decision Rule: Inspect the sample density plot. If the sample density plot is roughly bell-shaped and displays little skewness (following standard statistical rules of thumb), assume that a Normal Population Model is applicable to the overall population.

Sample Density versus Normal Population Model

Mathematical Properties of the Normal Density Curve

  • Area Under the Curve as Probability:
    • The total area under any density curve equals exactly 11 (or 100%100\%).
    • Sub-areas beneath the curve represent probabilities or population percentages (Area×100\text{Area} \times 100).
  • Core Geometric and Distributional Properties:
    • Symmetry: The curve is perfectly symmetric around its center.
    • Shape: Bell-shaped continuous distribution.
    • Modality: Unimodal (possesses exactly one central peak).
    • Central Tendency Identity: Mean=Median\text{Mean} = \text{Median}.
    • Symmetric Split: Exactly 50%50\% of the total population lies below the mean, and 50%50\% lies above the mean.

Properties of the Normal Model showing symmetry and equal 50% split

Calculating Probabilities Using the Normal Model

  • Finding Tail Probabilities:
    • To determine the percentage of a population meeting a threshold (e.g., the percent of all students learning about Santa after age 1010), calculate the area under the normal curve above that cutoff point (P(Y>10)P(Y > 10)).

Normal density plot with area above age 10 highlighted

  • Software Interface Instructions (Normal Probability Calculator):
    • Navigate to the Normal Probability Calculator section.
    • Select the target area type under Which tail? (e.g., Lower Tail, Upper Tail, or Between).
    • Specify the distribution parameters: Mean (μ\mu) and Standard Deviation (σ\sigma).
    • Enter the numeric boundary value corresponding to the sought probability area.

Stat 121 Normal Probability Calculator interface

  • Worked Examples (Given Normal Model with μ=8.391\mu = 8.391 and σ=2.684\sigma = 2.684):
    • Upper Tail Example: Probability that a student learns about Santa AFTER age 1010 (P(Y>10)P(Y > 10)):         P(Y>10)=0.2744(27.44%P(Y > 10) = 0.2744 \quad (27.44\%
    • Lower Tail Example: Percentage of students learning about Santa BEFORE age 66 (P(Y<6)P(Y < 6)):         P(Y<6)=0.1865(18.65%P(Y < 6) = 0.1865 \quad (18.65\%
    • Interval Example: Percentage of students learning about Santa BETWEEN ages 5.7075.707 and 11.07511.075 (P(5.707<Y<11.075)P(5.707 < Y < 11.075)):         P(5.707<Y<11.075)=0.6826(68.26%P(5.707 < Y < 11.075) = 0.6826 \quad (68.26\%

Calculating Percentiles and Tail Quantiles

  • Definition of Percentile: The p-thp\text{-th} percentile is the specific variable value such that p%p\% of the population lies below that value.
  • Software Interface Instructions (Percentiles from Normal Distribution):
    • Locate the bottom section of the normal probability tool titled Percentiles from Normal Distribution.
    • Select the percentile type under Which tail? (Lower Tail, Upper Tail, or Between).
    • Input the distribution mean (μ\mu) and standard deviation (σ\sigma).
    • Enter the target proportion percentile (a value between 00 and 11).

Stat 121 Percentiles from Normal Distribution interface

  • Worked Examples (Given Normal Model with μ=8.391\mu = 8.391 and σ=2.684\sigma = 2.684):
    • 75th Percentile: The 75th75\text{th} percentile age at which students learn about Santa is 10.201510.2015 years.
    • Theoretical Model Percentiles vs. Sample Quantiles:
      • The 75th75\text{th} percentile calculated from the normal model is not identical to Q3\text{Q3} calculated from sample data.
      • Reason: The normal population percentile is derived theoretically from the theoretical model distribution N(μ,σ)\mathcal{N}(\mu, \sigma), whereas Q3\text{Q3} is computed empirically from sample data.
    • Upper Tail Quantile Example: The age threshold above which the oldest 10%10\% of students find out about Santa (the 90th90\text{th} percentile) is 11.830711.8307 years.
    • Middle Area Percentile Strategy:
      • To find boundaries containing the middle C%C\% of a symmetric normal distribution, determine the lower boundary at the 100−C2th\frac{100 - C}{2}\text{th} percentile and the upper boundary at the 100+C2th\frac{100 + C}{2}\text{th} percentile.
      • Application for Middle 55%55\%:Lower Percentile=100−552=22.5th percentile=0.225\text{Lower Percentile} = \frac{100 - 55}{2} = 22.5\text{th percentile} = 0.225Upper Percentile=100+552=77.5th percentile=0.775\text{Upper Percentile} = \frac{100 + 55}{2} = 77.5\text{th percentile} = 0.775
      • Resulting Boundaries: The middle 55%55\% of BYU students find out about Santa between the ages of 6.36396.3639 and 10.418610.4186 years.

Middle 55% area under normal distribution curve

Standardization and Z-Scores

  • Rationale for Standardizing:

    • Observational data across different domains feature distinct units and scales:
      • Exam scores (00 to 100 points100\text{ points})
      • Tip amounts (dollars \, or percentage %\% of a bill)
      • Ages (years)
      • Liquid volumes (milliliters)
      • Reaction times (milliseconds)
    • Because scales differ, raw distances from an average cannot be directly compared (e.g., comparing a $3\$3 tip deviation directly to a 250 ms250\,\text{ms} reaction time deviation).
    • Standardizing converts variables from different scales onto a single standard scale, enabling:
      • Assessment of how extreme or unusual an individual observation is using a standard yardstick.
      • Direct comparison of observations originating from separate populations or different measurement units.
  • General Standardization Formula:Standardized Score=Estimate−MeanStandard Error\text{Standardized Score} = \frac{\text{Estimate} - \text{Mean}}{\text{Standard Error}}

    • Estimate: The observation, sample mean, or statistic being standardized.
    • Standard Error (SE): The typical degree to which the estimate varies around the mean.
  • Standardizing an Individual Observation (Z-scoreZ\text{-score}):

    • When standardizing a single observation YY from a Normal population, standard error equals the population standard deviation σ\sigma:         z=Y−μσz = \frac{Y - \mu}{\sigma}
    • Interpretation: A z-scorez\text{-score} represents the exact number of standard deviations an observation YY lies above or below the population mean μ\mu.
  • Single Observation Calculation Example:

    • Scenario: Calculate the z-scorez\text{-score} for a student who learns about Santa at age 1212, given μ=8.391\mu = 8.391 and \sigma = 2.684$.\n * *Calculation:*\n        z = \frac{12 - 8.391}{2.684} = 1.34\n * *Interpretation:* An age of 12isis1.34 standard deviations above the average age.\n\n* **Comparative Analysis Example (SAT vs. ACT Scores):**\n * *Scenario:* Jennie scores 600ontheSATMathexam.Geraldscoreson the SAT Math exam. Gerald scores27 on the ACT Math exam. Both tests evaluate the same underlying mathematical ability.\n * *Distribution Parameters:*\n * ACT Math Scores: \mathcal{N}(\mu = 18, \sigma = 6)\n * SAT Math Scores: \mathcal{N}(\mu = 500, \sigma = 100)\n * *Standardized Calculations:*\n        z_{\text{Jennie}} = \frac{600 - 500}{100} = 1.0\n        z_{\text{Gerald}} = \frac{27 - 18}{6} = 1.5\n * *Conclusion:* Gerald performed relatively higher than Jennie because his score is 1.5standarddeviationsaboveaverage,whereasJennie′sscoreisstandard deviations above average, whereas Jennie's score is1.0 standard deviation above average.\n\n# Summary of Core Terminology\n\n* **Statistical Model:** A mathematical formulation representing population distribution characteristics and shape.\n* **Normal Population Model:** Continuous probability distribution model defined by \muandand\sigma,denoted, denotedY \sim \mathcal{N}(\mu, \sigma).\n* **Greek vs. Roman Letters:** Convention designating population parameters with Greek symbols (\mu, \sigma)andsamplestatistics/datawithRomansymbols() and sample statistics/data with Roman symbols (s, Y, z).\n* **Probabilities from Normal Model:** Theoretical probability calculated as the area under the normal curve over a given interval.\n* **Percentiles from Normal Model:** Quantile value below which a given percentage p\% of the theoretical normal density lies.\n* **Standardizing:** Metric transformation z = \frac{Y - \mu}{\sigma}$$ quantifying how many standard deviations an observation falls from the mean.