6.1 Basic Statistics: Central Tendency & Dispersion

Basic Statistics: Central Tendency & Dispersion

Quick Answer: Basic statistics in Six Sigma characterize process behavior through two complementary dimensions: central tendency (location) and dispersion (variation). Central tendency is quantified by the mean (arithmetic average, sensitive to outliers), median (50th percentile, robust to skewness), and mode (most frequent value). Dispersion is measured by the range, variance (mean squared deviations), and standard deviation (spread in original units). To avoid underestimating population variation, sample standard deviation incorporates Bessel's correction by dividing by degrees of freedom ($n - 1$). Independent CSSYB study guide by OpenExamPrep.

Descriptive vs. Inferential Statistics

In the Measure phase of DMAIC, continuous improvement teams transition from qualitative problem definitions to rigorous, quantitative baselines. Statistical methods form the language of this phase, providing objective tools to evaluate process health, monitor variation, and identify improvement opportunities. Quality statistics fall into two major categories:

1. Descriptive Statistics: Techniques used to organize, summarize, tabulate, and visually present observed data. Descriptive statistics describe the specific characteristics of an observed data set—such as the average cycle time of 50 processed loan applications or the percentage of defective welds in a test batch—without extending conclusions beyond the collected sample.
2. Inferential Statistics: Techniques that utilize sample data to draw probabilistic conclusions, estimate parameters, and test hypotheses regarding a broader, unobserved population or future process output. Because inspecting 100% of a production run is frequently impractical, expensive, or destructive, Six Sigma relies on inferential statistics to infer population truth from representative samples.

While descriptive statistics dominate the early Measure phase to establish baseline stability, inferential methods are deployed extensively in the Analyze and Improve phases to validate root causes and test process changes.

---

Population Parameters vs. Sample Statistics

In statistical quality control, a fundamental distinction exists between the entire group of interest (the population) and the subset selected for analysis (the sample):

• Population: The complete collection of all possible units, events, components, or transactions that share a defined characteristic. For example, every medical claim processed by an insurance company in 2026 constitutes a population. Numerical values describing an entire population are called parameters and are traditionally denoted by Greek letters.
• Sample: A representative subset of units drawn from the population. Numerical values computed from sample data are called statistics and are denoted by Roman letters.

In operational environments, true population parameters are rarely known because manufacturing processes generate ongoing streams of output over time. Instead, Yellow Belts collect samples to calculate sample statistics ($\bar{x}, s$), which serve as point estimators for the true underlying population parameters ($\mu, \sigma$).

---

Measures of Central Tendency (Location)

Measures of central tendency identify the center or typical value of a probability distribution or data set. In Six Sigma, understanding central tendency allows practitioners to evaluate whether a process is centered on customer target specifications.

1. The Arithmetic Mean ($\bar{x}$ or $\mu$)

The arithmetic mean is the most common measure of central tendency. It represents the mathematical center of gravity of a data set, calculated by summing all individual values and dividing by the total count:

$$\bar{x} = \frac{\sum_{i=1}^{n} x_i}{n}$$

• Strengths: Incorporates all numerical values in the data set; possesses well-understood mathematical properties essential for inferential statistics and control charting (such as $\bar{X}$-charts).
• Weaknesses: Highly sensitive to extreme outliers and asymmetric skewness. A single extreme data point (e.g., an invoice delayed by 90 days due to a system glitch) pulls the mean toward the tail, presenting a distorted impression of typical performance.

2. The Median ($\tilde{x}$)

The median is the middle value of an ordered data set, dividing the ranked observations into two equal halves (the 50th percentile):

• When sample size $n$ is odd, the median is the single observation at position $(n + 1) / 2$.
• When sample size $n$ is even, the median is the arithmetic mean of the two middle observations at positions $n / 2$ and $(n / 2) + 1$.
• Strengths: Robust and resistant to extreme outliers and skewed data. In cycle-time analyses, customer service queues, and income metrics, the median often provides a more realistic representation of typical customer experience than the mean.
• Weaknesses: Does not utilize every data value's exact magnitude; less mathematically amenable to standard parametric hypothesis tests.

3. The Mode

The mode is the value that occurs with the greatest frequency in a data set:

• A distribution may have a single peak (unimodal), two distinct peaks (bimodal), multiple peaks (multimodal), or no repeating values at all.
• Strengths: The only measure of central tendency applicable to nominal (categorical) data (e.g., determining the most common customer complaint type or frequent defect code).
• Process Insight: The emergence of a bimodal distribution in continuous process data often signals that two distinct processes or conditions have been inadvertently mixed together—such as parts produced on two different machine tooling fixtures, raw materials from two alternate suppliers, or different operating behaviors across day and night shifts.

---

Measures of Dispersion (Spread and Variation)

In Six Sigma, assessing central tendency alone is inadequate. As the classic statistical cautionary tale notes, a person with one foot in freezing ice water and the other in boiling water may be "comfortable on average," yet is in severe physical distress. In quality engineering, variation is the primary driver of customer dissatisfaction and process defects. Two processes can share an identical mean of 50.0 units, yet one process may consistently operate between 49.5 and 50.5 (capable), while the other fluctuates wildly between 30.0 and 70.0 (generating massive scrap).

1. The Range ($R$)

The range is the simplest measure of dispersion, representing the arithmetic difference between the largest and smallest values in a data set:

$$R = x_{\max} - x_{\min}$$

• Strengths: Rapid and intuitive to calculate on the shop floor without software; forms the mathematical basis for range charts ($R$-charts) in Statistical Process Control for small subgroups ($n \le 10$).
• Weaknesses: Utilizes only two extreme observations, discarding all intermediate data; increases systematically as sample size grows and is vulnerable to single anomalous outliers.

2. Variance ($\sigma^2$ and $s^2$)

Variance measures the average squared deviation of individual data points from their arithmetic mean. Squared deviations ensure that negative deviations (values below the mean) do not cancel out positive deviations (values above the mean):

• Population Variance:

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

• Sample Variance:

$$s^2 = \frac{\sum_{i=1}^{n} (x_i - \bar{x})^2}{n - 1}$$

Why Divide by $n - 1$? (Bessel's Correction)

A frequent question among quality students is why sample variance divides by $(n - 1)$ rather than $n$. When drawing a sample from a broader population, the individual data points cluster more tightly around their own sample mean ($\bar{x}$) than around the true, unknown population mean ($\mu$).

Consequently, calculating sample variance with a denominator of $n$ produces a biased estimator that systematically underestimates the true variation present in the population. Dividing by the degrees of freedom, $n - 1$, mathematically corrects this downward bias, producing an unbiased estimator of population variance. As sample size $n$ increases, the difference between $n$ and $n - 1$ becomes negligible.

3. Standard Deviation ($\sigma$ and $s$)

While variance is mathematically indispensable for statistical proofs and Analysis of Variance (ANOVA), it suffers from a significant practical drawback: it is expressed in squared units of measurement (e.g., minutes squared, dollars squared, square millimeters). Squared units cannot be directly compared against linear product tolerances or customer specification limits.

To restore dispersion to the original physical units of the process, practitioners take the positive square root of the variance:

• Population Standard Deviation: $\sigma = \sqrt{\sigma^2} = \sqrt{\frac{\sum (x_i - \mu)^2}{N}}$
• Sample Standard Deviation: $s = \sqrt{s^2} = \sqrt{\frac{\sum (x_i - \bar{x})^2}{n - 1}}$

Standard deviation serves as the foundational metric of Six Sigma methodology. It directly determines process capability ratios ($C_p, C_{pk}$), defines statistical control chart boundaries (placed at $\pm 3\sigma$), and quantifies the defect rate under normal distribution assumptions.

---

Step-by-Step Worked Calculation Example

A Six Sigma Yellow Belt on a transactional banking team collects a random sample of $n = 8$ loan application processing times (recorded in minutes) to establish baseline performance:

$$\{24, 28, 22, 25, 24, 32, 26, 27\}$$

Step 1: Arrange the Data in Ascending Order

$$\{22, 24, 24, 25, 26, 27, 28, 32\}$$

Step 2: Compute the Sample Mean ($\bar{x}$)

$$\sum_{i=1}^{8} x_i = 22 + 24 + 24 + 25 + 26 + 27 + 28 + 32 = 208$$ $$\bar{x} = \frac{208}{8} = 26.0 \text{ minutes}$$

Step 3: Identify the Median

Because the sample size $n = 8$ is even, the median is the average of the 4th and 5th ordered values:

• 4th value $= 25$
• 5th value $= 26$
• $\text{Median} = \frac{25 + 26}{2} = 25.5 \text{ minutes}$

Step 4: Identify the Mode

The value 24 appears twice, while all other observations appear once. Therefore, the distribution is unimodal with $\text{Mode} = 24.0 \text{ minutes}$.

Step 5: Compute the Range ($R$)

$$R = x_{\max} - x_{\min} = 32 - 22 = 10.0 \text{ minutes}$$

Step 6: Construct the Deviation and Sum of Squares Table

(Note: The algebraic sum of deviations from the arithmetic mean always equals exactly zero, serving as an internal arithmetic check.)

Step 7: Calculate Sample Variance ($s^2$)

Applying Bessel's correction with $n - 1 = 8 - 1 = 7$ degrees of freedom:

$$s^2 = \frac{\sum (x_i - \bar{x})^2}{n - 1} = \frac{66.0}{7} \approx 9.4286 \text{ minutes}^2$$

Step 8: Calculate Sample Standard Deviation ($s$)

Taking the square root of sample variance:

$$s = \sqrt{9.4286} \approx 3.07 \text{ minutes}$$

Process Interpretation: The loan processing operation exhibits an average cycle time of 26.0 minutes with a standard deviation of 3.07 minutes. Under normal distribution assumptions, approximately 99.73% of all applications processed under stable conditions will finish within the $\bar{x} \pm 3s$ window: $26.0 \pm 3(3.07) = [16.79, 35.21]$ minutes. If customer Service Level Agreements (SLAs) demand all approvals within 30 minutes, this baseline analysis immediately alerts the team that the upper process tail ($35.21$ minutes) causes non-compliance, pinpointing the focus for subsequent DMAIC phases.