Average & Dataset

When conducting scientific investigations, financial simulations, or structural engineering audits, researchers must look past standard descriptive summaries to fully grasp the underlying frequency dynamics. A simple average represents only the absolute balance point of a dataset; it reveals nothing of the dataset's internal structure, skewness, or standard deviation dispersion. By exploring descriptive variables in parallel—specifically arithmetic, geometric, and harmonic means, coupled with variance spreads—data analysts can extract actionable structural models.

---

📊 The Hierarchy of Central Tendency: Arithmetic, Geometric, and Harmonic Averages

The choice of statistical average is deeply dependent on the dimensional properties of the dataset. The **Arithmetic Mean** ($\mu$) is the most ubiquitous model, computed by dividing the sum of observations by the count ($N$). However, because the arithmetic mean scales linearly, it is exceptionally sensitive to extreme outlier spikes, which can distort the resolved balance point.

When modeling compounding interest growth, investment portfolio returns, or biological population trajectories, researchers must instead utilize the **Geometric Mean** ($G$). Because growth acts multiplicatively rather than additively, the geometric mean calculates the actual compounded rate of change. The geometric mean of a dataset is always less than or equal to the arithmetic mean: $G \le \mu$.

For calculations involving rates (such as velocity, density, or fuel consumption), the **Harmonic Mean** ($H$) represents the correct physical metric. If a vehicle traverses a distance at $40 \text{ km/h}$ and returns at $60 \text{ km/h}$, the average velocity is not $50 \text{ km/h}$ (the arithmetic mean), but rather $48 \text{ km/h}$ (the harmonic mean), as speed is inversely proportional to time spent.

$$\text{The Pythagorean Means relationship holds true: } H \le G \le \mu$$

---

📈 Measuring Dispersion: Standard Deviation & Variance

While the mean solves the location of a dataset's center, **Variance** ($\sigma^2$) and **Standard Deviation** ($\sigma$) quantify the dispersion of values around that center. Variance represents the average squared distance from the mean, preventing positive and negative offsets from canceling each other out:

$$\sigma^2 = \frac{1}{N}\sum_{i=1}^N (x_i - \mu)^2$$

Because variance is expressed in squared units (which distorts physical interpretability), taking the square root yields the **Standard Deviation** ($\sigma$), returning the dispersion variable back to the dataset's original physical dimension.

Under Chebyshev's Inequality and Normal Distribution models (assuming a symmetrical bell curve), approximately **68.27%** of all data points fall within one standard deviation ($\mu \pm \sigma$) of the mean, **95.45%** within two standard deviations ($\mu \pm 2\sigma$), and **99.73%** within three standard deviations ($\mu \pm 3\sigma$). High standard deviations indicate broad data scatter, whereas low standard deviations represent closely grouped values.

In financial risk assessment (such as evaluating exchange rates like the Icelandic Króna ISK), the **Coefficient of Variation** ($\text{CV} = \sigma / \mu$) normalizes standard deviation relative to the mean, allowing analysts to compare volatility profiles across disparate asset classes regardless of scale.

---

📐 Skewness, Modality, and Outlier Resistance (Median vs. Mean)

In asymmetrical distributions, relying strictly on the arithmetic mean can lead to highly biased conclusions. When a dataset contains extreme outliers—such as corporate income datasets where a handful of high earners distort the general average, or real estate values in the Greater Reykjavík area—the **Median** represents a far more robust measure of central tendency. The Median represents the exact 50th percentile mark: the value dividing the upper half of sorted observations from the lower half. Because the median relies on positional indexing rather than additive magnitudes, a single extreme outlier will not shift its value.

Understanding the relationships between **Mean**, **Median**, and **Mode** (the most frequently occurring value in the dataset) reveals the **Skewness** of the distribution. In a perfectly symmetrical normal distribution (a bell-shaped Gaussian curve), the mean, median, and mode are mathematically identical. In a *positively skewed* distribution (skewed to the right), the tail extends towards larger positive values, pulling the mean higher than the median ($\text{Mean} > \text{Median} > \text{Mode}$). Conversely, in a *negatively skewed* distribution (skewed to the left), the tail extends towards smaller or negative values, dragging the mean below the median ($\text{Mean} < \text{Median} < \text{Mode}$). Tracking these three values together gives data scientists a clear diagnostic view of distribution asymmetry.