7.1 Descriptive Statistics, Data Distributions & Visual Displays

7.1 Descriptive Statistics, Data Distributions & Visual Displays

Descriptive statistics forms the bedrock of exploratory data analysis (EDA) in secondary mathematics. For candidates preparing for the NBPTS Adolescence and Young Adulthood (AYA) Mathematics certification, mastery extends beyond mechanical calculation of summary numbers. Accomplished teachers must synthesize conceptual understandings of center and dispersion, evaluate the resistance of statistical estimators under distribution skewness, justify degrees of freedom in variance formulas, and diagnose persistent student misconceptions when interpreting graphical representations of univariate data.

---

1. Measures of Center: Mean, Median, and Mode

A measure of central tendency identifies a single representative value that summarizes the location of a distribution's center.

The Arithmetic Mean

For a finite population of size $N$, the population mean is denoted by the Greek letter $\mu$: $\mu = \frac{1}{N} \sum_{i=1}^{N} x_i$ For a sample of size $n$, the sample mean is denoted by $\bar{x}$ (x-bar): $\bar{x} = \frac{1}{n} \sum_{i=1}^{n} x_i$

Mathematically, the arithmetic mean acts as the physical center of gravity or balance point of the data. If each data point is placed on a line with equal unit mass, the distribution balances perfectly at $\bar{x}$. Consequently, the sum of deviations from the mean is always zero: $\sum_{i=1}^{n} (x_i - \bar{x}) = \sum_{i=1}^{n} x_i - \sum_{i=1}^{n} \bar{x} = n\bar{x} - n\bar{x} = 0$

The Median

The median (denoted $\text{Med}$ or $Q_2$) is the 50th percentile of an ordered distribution. It divides an ordered dataset into two halves of equal cardinality:

1. Arrange $n$ observations in ascending order: $x_{(1)} \le x_{(2)} \le \dots \le x_{(n)}$.
2. If $n$ is odd, the median is the unique middle observation at position $\frac{n + 1}{2}$: $\text{Med} = x_{\left(\frac{n+1}{2}\right)}$
3. If $n$ is even, the median is the arithmetic mean of the two central observations at positions $\frac{n}{2}$ and $\frac{n}{2} + 1$: $\text{Med} = \frac{x_{(n/2)} + x_{(n/2 + 1)}}{2}$

The Mode

The mode is the most frequently occurring observation in a dataset. A distribution may be unimodal (one clear peak), bimodal (two distinct local peaks), multimodal, or have no mode if all values occur with equal frequency. The mode is the only measure of central tendency applicable to nominal categorical data.

Statistical Resistance

A statistic is termed resistant (or robust) if extreme outliers or substantial skewness in the tails have minimal impact on its numerical value:

• The Median is resistant: Changing the value of the single largest observation $x_{(n)}$ to an arbitrarily large number does not alter the median, because the middle positional ordering remains invariant.
• The Mean is non-resistant: Because every single value enters the sum $\sum x_i$, an extreme value pulls $\bar{x}$ markedly toward it.

+-----------------------------------------------------------------------------------+
|                         CENTER MEASURE COMPARISON MATRIX                          |
+-----------+----------------------+--------------------+---------------------------+
| Measure   | Sensitivity to Tails | Resistance Level   | Optimal Distributional Use|
+-----------+----------------------+--------------------+---------------------------+
| Mean      | High (pulled by tail)| Non-Resistant      | Symmetric, Bell-Shaped    |
| Median    | None (position-based)| Highly Resistant   | Skewed or Outlier-Laden   |
| Mode      | None (frequency-only)| Locally Resistant  | Categorical / Peak Finding|
+-----------+----------------------+--------------------+---------------------------+

---

2. Measures of Spread: Range, IQR, Variance, and Standard Deviation

Describing a distribution solely by its center is fundamentally incomplete. Measures of dispersion quantify the degree of variability, diversity, or scatter among observations.

Range

The range is the difference between the maximum and minimum observations: $\text{Range} = x_{(n)} - x_{(1)} = \text{Max} - \text{Min}$ Because it relies entirely on the two extreme values, the range is highly non-resistant and provides no information about how observations are distributed between the extremes.

Interquartile Range (IQR)

The interquartile range measures the spread of the central 50% of the data. It is defined as the difference between the third quartile ($Q_3$, 75th percentile) and the first quartile ($Q_1$, 25th percentile): $\text{IQR} = Q_3 - Q_1$ Like the median, the IQR is a resistant measure of spread. Trimming or altering values outside the first and third quartiles leaves the IQR completely unchanged.

Variance and Standard Deviation

Variance measures the mean squared deviation of data points from their arithmetic center.

Population Variance and Standard Deviation

For a population of size $N$ with mean $\mu$: $\sigma^2 = \frac{1}{N} \sum_{i=1}^{N} (x_i - \mu)^2, \qquad \sigma = \sqrt{\frac{1}{N} \sum_{i=1}^{N} (x_i - \mu)^2}$

Sample Variance and Bessel's Correction

For a sample of size $n$ with sample mean $\bar{x}$, the sample variance $s^2$ is defined with Bessel's correction, dividing by $n - 1$ rather than $n$: $s^2 = \frac{1}{n - 1} \sum_{i=1}^{n} (x_i - \bar{x})^2, \qquad s = \sqrt{\frac{1}{n - 1} \sum_{i=1}^{n} (x_i - \bar{x})^2}$

[!IMPORTANT] Mathematical Derivation of Bessel's Correction ($n - 1$ Degrees of Freedom): Why divide by $n - 1$? When estimating $\sigma^2$ using the sample deviations $(x_i - \bar{x})$, the sample mean $\bar{x}$ is itself calculated from the sample. This enforces the linear constraint $\sum_{i=1}^{n} (x_i - \bar{x}) = 0$. Once $n - 1$ deviations are known, the $n$-th deviation is completely determined. Hence, there are only $n - 1$ independent pieces of information (degrees of freedom).

More critically, observations $x_i$ are closer on average to their own sample mean $\bar{x}$ than to the true population mean $\mu$. Specifically, it can be proven algebraically that: $\sum_{i=1}^{n} (x_i - \mu)^2 = \sum_{i=1}^{n} (x_i - \bar{x})^2 + n(\bar{x} - \mu)^2$ Taking mathematical expectations yields $E\left[\frac{1}{n}\sum_{i=1}^{n}(x_i - \bar{x})^2\right] = \frac{n - 1}{n}\sigma^2$. Dividing by $n$ produces a statistic that systematically underestimates the population variance (a biased estimator). Dividing by $n - 1$ compensates exactly for this downward bias, ensuring that $E[s^2] = \sigma^2$, making $s^2$ an unbiased estimator of $\sigma^2$.

Both $s^2$ and $s$ are non-resistant measures of spread; extreme values produce large squared deviations $(x_i - \bar{x})^2$ that inflate standard deviation substantially.

---

3. Visual Displays: Stemplots, Histograms, and Boxplots

Visual representations permit rapid evaluation of distribution shape, center, spread, and anomalies.

Stemplots (Stem-and-Leaf Displays)

A stemplot separates each numerical value into a "stem" (leading digits) and a "leaf" (final trailing digit).

• Key Advantages: Retains raw data values; organizes observations into ordered rank; reveals modality and skewness simultaneously.
• Techniques: Splitting stems (e.g., placing leaves 0–4 on the first stem and 5–9 on the second) prevents crowding; back-to-back stemplots allow direct visual comparison of two independent distributions sharing identical stem units.

Histograms

A histogram bins continuous quantitative data into contiguous intervals (classes) along the horizontal axis, plotting bar heights proportional to frequency, relative frequency, or probability density.

• Bin Width: The choice of bin width critically influences the visual impression: excessively wide bins obscure modality, while excessively narrow bins introduce noise.
• The Area Principle: In density histograms, the area of each bar represents the relative frequency of that class: $\text{Density} = \frac{\text{Relative Frequency}}{\text{Bin Width}}$ When bin widths are equal, bar height is directly proportional to frequency; when bin widths are unequal, heights must represent density to prevent visual distortion.

Boxplots & Tukey's 1.5 × IQR Outlier Rule

A boxplot (box-and-whisker plot) displays the Five-Number Summary: $\text{Five-Number Summary} = \{ \text{Min}, Q_1, \text{Median}, Q_3, \text{Max} \}$

                Q1 - 1.5*IQR                     Q3 + 1.5*IQR
                Lower Fence                      Upper Fence
                     |                                |
  Outlier            v                                v               Outlier
    *        |-------[==============|=================]-------|          *
            Min     Q1            Median             Q3      Max
                     |<------------ IQR ------------->|

The 1.5 × IQR Outlier Rule (Tukey's Fences)

John Tukey established standard analytical thresholds to flag potential outliers:

1. Calculate $\text{IQR} = Q_3 - Q_1$.
2. Compute inner fences: $\text{Lower Inner Fence} = Q_1 - 1.5 \times \text{IQR}$ $\text{Upper Inner Fence} = Q_3 + 1.5 \times \text{IQR}$
3. In a modified boxplot, whiskers extend only to the most extreme data values that fall inside the inner fences. Any data point falling strictly outside the inner fences is plotted as an isolated symbol (e.g., dot or asterisk) representing a mild outlier.
4. Extreme outliers are values that fall outside outer fences defined by $Q_1 - 3.0 \times \text{IQR}$ or $Q_3 + 3.0 \times \text{IQR}$.

---

4. Distribution Shape, Skewness & Resistance

The relationship between measures of center and spread depends heavily on distribution symmetry.

Why Mean Follows the Tail

In a right-skewed distribution, the large values in the right tail produce large positive deviations $(x_i - \bar{x})$. Because the sum of deviations must balance to zero, the mean must shift to the right to counterbalance these extreme values. The median, depending only on the positional rank of ordered data, remains stationary near the bulk of the observations.

---

5. Linear Transformations of Univariate Data

In scientific measurement and secondary classroom grading, data are frequently subjected to linear transformations of the form: $y_i = a x_i + b \quad (a \ne 0)$

Mathematical Impact on Measures of Center and Position

Adding a constant $b$ shifts every observation by $b$ units along the number line. Multiplying by $a$ rescales distances by factor $a$. Therefore, all measures of center and location (mean, median, mode, quartiles, percentiles) are affected by both multiplication and addition: $\bar{y} = a\bar{x} + b, \qquad \text{Med}_y = a\text{Med}_x + b, \qquad Q_{1,y} = aQ_{1,x} + b, \qquad Q_{3,y} = aQ_{3,x} + b$

Mathematical Impact on Measures of Spread

Shifting an entire distribution by $+b$ does not alter the relative distance between any two data points: $(y_i - y_j) = (ax_i + b) - (ax_j + b) = a(x_i - x_j)$. Consequently, measures of spread (range, IQR, standard deviation) are completely unaffected by the additive constant $b$, scaling solely by the absolute value $|a|$: $\text{Range}_y = |a| \times \text{Range}_x$ $\text{IQR}_y = |a| \times \text{IQR}_x$ $s_y = |a| \times s_x$ $s_y^2 = a^2 \times s_x^2$

[!NOTE] Standardizing Data ($z$-score Transformation): Standardizing an observation creates a dimensionless measure of relative standing: $z_i = \frac{x_i - \bar{x}}{s} = \left(\frac{1}{s}\right)x_i - \frac{\bar{x}}{s}$. Here $a = 1/s$ and $b = -\bar{x}/s$. Under this transformation, the standardized mean becomes $\bar{z} = 0$ and the standardized standard deviation becomes $s_z = 1$.

---

6. Secondary Student Misconceptions & Pedagogical Interventions

Accomplished mathematics teachers anticipate common student reasoning errors when teaching univariate data analysis:

Misconception 1: Misinterpreting Boxplot Span as Frequency or Data Count

Students frequently examine a modified boxplot with an elongated whisker or wide quartile box and assert that "there are more data points in this section because the interval is longer."

• Pedagogical Intervention: Require students to construct dotplots directly above corresponding boxplots. Emphasize that every segment of a boxplot—from minimum to $Q_1$, $Q_1$ to median, median to $Q_3$, and $Q_3$ to maximum—contains approximately 25% of the data. A wider span does not indicate more data; it indicates that the same proportion (25%) of observations is more dispersed (lower density).

Misconception 2: Confusing Histogram Bar Height with Data Values

When viewing a histogram of test scores, secondary students often confuse the vertical axis with the value of the variable rather than the count or frequency of observations, mistaking the display for a bar chart or time series.

• Pedagogical Intervention: Have students manually construct a frequency table before rendering the histogram. Guide them through the two-step translation: (1) partition the quantitative continuum into disjoint class intervals along the horizontal axis, and (2) construct vertical bars whose heights record the tally of occurrences within each interval.

Misconception 3: Believing Standard Deviation Can Be Negative or Increased by Shifting

Students often calculate a negative sample standard deviation or believe that adding 10 bonus points to all student exam scores will increase the class standard deviation by 10 points.

• Pedagogical Intervention: Emphasize the algebraic definition $s = \sqrt{\frac{\sum (x_i - \bar{x})^2}{n - 1}}$. Because squared deviations are non-negative, variance and standard deviation satisfy $s^2 \ge 0$ and $s \ge 0$ universally ($s = 0$ if and only if all observations are identical). Use dynamic geometry software to slide an entire dataset along the number line, demonstrating visually that rigid translations preserve all pairwise distances, keeping standard deviation invariant.