6.1 Descriptive Statistics & Data Summarization

6.1 Descriptive Statistics & Data Summarization

Descriptive statistics provide the quantitative foundation for quality engineering. Before executing statistical process control (SPC), design of experiments (DOE), or hypothesis testing, a quality engineer must summarize and interpret the baseline characteristics of collected process data. Data summarization relies on three core dimensions: central tendency (where the center lies), dispersion (how spread out the values are), and shape (the symmetry and peakedness of the distribution).

---

Measures of Central Tendency

Central tendency describes the central or typical value around which data points cluster. The three primary metrics tested on the ASQ CQE exam are the arithmetic mean, median, and mode.

1. Arithmetic Mean

The sample mean ($\bar{x}$) is the arithmetic average of $n$ observations, whereas the population mean ($\mu$) represents the average of all $N$ units in the population.

$\bar{x} = \frac{1}{n} \sum_{i=1}^{n} x_i, \quad \mu = \frac{1}{N} \sum_{i=1}^{N} X_i$

• Properties: Incorporates every data point; mathematically tractible for statistical inference.
• Limitations: Highly sensitive to extreme values or outliers. A single severe measurement error distorts $\bar{x}$.

2. Median

The median ($\text{me}$ or $\tilde{x}$) is the middle observation when data are arranged in ascending order. If sample size $n$ is odd, it is the value at position $(n+1)/2$. If $n$ is even, it is the mean of the two central values at positions $n/2$ and $(n/2)+1$.

• Properties: Highly robust against outliers and extreme skewness. Preferred in failure-rate analysis and income data.

3. Mode

The mode is the most frequently occurring value in the dataset. A distribution may be unimodal, bimodal, multimodal, or have no mode.

• Properties: Useful for attribute data (e.g., most common defect type) and identifying mixture distributions.

---

Measures of Dispersion & Variability

Variability is the root cause of quality degradation. Two processes can share an identical mean of $10.0\text{ mm}$, but a process with wide variation will produce defective parts outside engineering specification limits.

1. Range ($R$)

The difference between the maximum and minimum observed values:

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

While quick to compute and standard in small-sample control charts ($X$-bar and $R$ charts where $n \le 10$), the range utilizes only two data points and underestimates true population dispersion for large samples.

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

Variance measures the mean squared deviation of data points from their average. Population variance $\sigma^2$ divides by population size $N$:

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

Sample variance $s^2$ uses Bessel's correction ($n-1$ in the denominator) to eliminate downward bias caused by using sample mean $\bar{x}$ instead of population mean $\mu$:

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

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

The square root of variance restores the metric to the original units of measurement:

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

4. Coefficient of Variation ($CV$)

The Coefficient of Variation is a dimensionless measure of relative variability, expressed as a percentage:

$CV = \left( \frac{s}{\bar{x}} \right) \times 100\%$

$CV$ allows quality engineers to compare variability between two processes with vastly different means or different units of measurement (e.g., weight in grams vs. length in millimeters).

---

Distribution Shape: Skewness & Kurtosis

Understanding data symmetry and tail weights ensures appropriate selection of control charts and capability indices ($C_p, C_{pk}$ vs. non-normal equivalents).

Skewness

Skewness quantifies the degree of asymmetry of a distribution around its mean:

Kurtosis

Kurtosis measures the relative peakedness or tail heaviness of a probability distribution compared to a normal distribution:

• Mesokurtic ($g_2 = 0$): Normal distribution curve standard.
• Leptokurtic ($g_2 > 0$): Sharp central peak with heavy, fat tails (higher probability of extreme outliers).
• Platykurtic ($g_2 < 0$): Flat peak with thin tails (more uniform distribution).

---

Summary Table of Descriptive Formulas

---

Comprehensive Worked Numerical Example

Scenario: A quality engineer collects $n = 10$ shaft diameter measurements (in millimeters) from a CNC lathe turning process:
[10.02, 10.05, 9.98, 10.01, 10.04, 10.00, 10.06, 9.99, 10.03, 10.02]

Calculate the sample mean, median, mode, range, sample variance, sample standard deviation, and coefficient of variation.

Step-by-Step Solution:

1. Sort Data Ascending:
   9.98, 9.99, 10.00, 10.01, 10.02, 10.02, 10.03, 10.04, 10.05, 10.06

2. Sample Mean ($\bar{x}$):
   $\sum x_i = 100.20\text{ mm}$ $\bar{x} = \frac{100.20}{10} = 10.020\text{ mm}$

3. Median ($\text{me}$):
   Since $n = 10$ is even, average 5th ($10.02$) and 6th ($10.02$) values:
   $\text{Median} = \frac{10.02 + 10.02}{2} = 10.020\text{ mm}$

4. Mode:
   The value $10.02$ appears twice. $\text{Mode} = 10.020\text{ mm}$.

5. Range ($R$):
   $R = 10.06 - 9.98 = 0.080\text{ mm}$

6. Sum of Squared Deviations ($\sum (x_i - \bar{x})^2$):
   Direct deviations from $\bar{x}=10.02$: $0$, $0.0009$, $0.0016$, $0.0001$, $0.0004$, $0.0004$, $0.0016$, $0.0009$, $0.0001$, $0$ sum to $0.0060$. $\sum x_i^2 = 1004.0100$ $\sum (x_i - \bar{x})^2 = 1004.0100 - \frac{(100.20)^2}{10} = 1004.0100 - 1004.0040 = 0.006000\text{ mm}^2$

7. Sample Variance ($s^2$):
   $s^2 = \frac{0.006000}{10 - 1} = \frac{0.006000}{9} = 0.0006667\text{ mm}^2$

8. Sample Standard Deviation ($s$):
   $s = \sqrt{0.0006667} = 0.02582\text{ mm}$

9. Coefficient of Variation ($CV$):
   $CV = \left( \frac{0.02582}{10.020} \right) \times 100\% = 0.2577\%$

---

CQE Exam Tips & Strategy

• Watch the Denominator: Always check whether a problem specifies a sample ($n-1$) or a population ($N$). Default to sample formulas ($s^2, s$) for quality inspection problems.
• Identify Skewness Quickly: On ASQ exam questions, if $\text{Mean} > \text{Median}$, the distribution is right-skewed (positive skew). If $\text{Mean} < \text{Median}$, it is left-skewed (negative skew).
• CV Applications: Use $CV$ when comparing process variability across different scales (e.g., comparing micro-meter tolerances vs. meter structural beam lengths).