10.3 Statistical Concepts in Assessment: Mean, Median, Standard Deviation, Percentiles

Four Scales of Measurement in Assessment

Educational statistics begins with understanding the four scales of measurement (NOIR framework): Nominal, Ordinal, Interval, and Ratio.

1. Nominal Scale: Categorical classification without numerical order or magnitude. Numbers serve merely as labels (e.g., coding Academic Track as 1, TVL Track as 2, Sports Track as 3).
2. Ordinal Scale: Categorical data with a distinct rank order, but unequal or unknown intervals between ranks (e.g., class rankings: 1st, 2nd, 3rd place; percentile ranks; Likert scale responses).
3. Interval Scale: Ordered data with equal numerical intervals between points, but containing an arbitrary (zero point does not mean absence) (e.g., temperature in Celsius, standardized IQ scores). A score of zero on an IQ test does not mean zero intelligence.
4. Ratio Scale: Ordered data with equal intervals and an absolute, true zero point representing complete absence of the property (e.g., height, weight, execution speed in seconds, raw score count).

---

Measures of Central Tendency

Measures of central tendency describe the single central or typical value around which raw test scores cluster.

1. The Arithmetic Mean ($\bar{X}$)

The mean is the arithmetic average of all scores in a distribution:

$$\bar{X} = \frac{\sum X}{N}$$

• Properties: Uses every score in the dataset. Highly sensitive to extreme scores (outliers).
• Best Use: Symmetrical, unskewed distributions of interval or ratio data.

2. The Median ($Md$)

The median is the physical middle score when data are arranged in ascending or descending order (the 50th percentile).

• Properties: Dividing score line where 50% fall above and 50% fall below. Insensitive to extreme outliers.
• Best Use: Skewed distributions or ordinal ranked data.

3. The Mode ($Mo$)

The mode is the most frequently occurring score in a distribution.

• Properties: May be unimodal (one mode), bimodal (two modes), or multimodal. Unstable across samples.
• Best Use: Nominal data or identifying the most popular choice.

---

Measures of Variability (Dispersion)

Variability describes how spread out or clustered raw scores are around the central tendency.

• Range: The difference between the highest and lowest score ($H - L + 1$). Extremely sensitive to extreme values.
• Variance ($s^2$): The average squared deviation of scores from the mean: $$s^2 = \frac{\sum (X - \bar{X})^2}{N - 1}$$
• Standard Deviation ($s$ or $\sigma$): The positive square root of the variance ($s = \sqrt{s^2}$). It measures the average distance of scores from the mean in original score units.
  • Small Standard Deviation: Indicates low score dispersion. Student performance is homogeneous (clustered tightly around the mean).
  • Large Standard Deviation: Indicates high score dispersion. Student performance is heterogeneous (widely scattered).

---

Distribution Shapes and Skewness

The relationship between the Mean, Median, and Mode determines the symmetry and shape of a frequency distribution curve.

                 DISTRIBUTION CURVE SKEWNESS

      NORMAL                 POSITIVE SKEW             NEGATIVE SKEW
   (Symmetrical)             (Tail to Right)           (Tail to Left)
        ▲                         ▲                         ▲
       ╱ ╲                       ╱ ╲                       ╱ ╲
      ╱   ╲                     ╱   ╲──┐               ┌──╱   ╲
     ╱     ╲                   ╱       ╲               ╱       ╲
   ─┴───────┴─               ─┴─────────┴─           ─┴─────────┴─
   Mean=Md=Mode              Mode < Md < Mean        Mean < Md < Mode
   (Normal Test)             (Difficult Test)         (Easy Test)

1. Normal Distribution (Bell Curve)

• Symmetrical, bell-shaped curve where $\text{Mean} = \text{Median} = \text{Mode}$.
• Empirical Rule (68-95-99.7 Rule):
  • $\pm 1\sigma$ encompasses approximately 68.26% of all scores.
  • $\pm 2\sigma$ encompasses approximately 95.44% of all scores.
  • $\pm 3\sigma$ encompasses approximately 99.72% of all scores.

2. Positively Skewed Distribution

• The tail of the curve extends toward the right (higher values).
• Directional relationship: $\text{Mean} > \text{Median} > \text{Mode}$.
• Pedagogical Meaning: The test was DIFFICULT. Most students scored low (clustering on the left), while only a few high-achieving students pulled the mean to the right.

3. Negatively Skewed Distribution

• The tail of the curve extends toward the left (lower values).
• Directional relationship: $\text{Mode} > \text{Median} > \text{Mean}$.
• Pedagogical Meaning: The test was EASY. Most students scored high (clustering on the right), while a few low-performing students pulled the mean to the left.

---

Standardized Scores: Z-Scores and T-Scores

Standardized scores convert raw scores into relative position metrics based on standard deviation units.

1. Z-Score

Expresses how many standard deviations a raw score ($X$) lies above or below the mean ($\bar{X}$):

$$z = \frac{X - \bar{X}}{s}$$

• A $z$-score of $0$ equals the mean score.
• A positive $z$-score indicates performance above the mean; a negative $z$-score indicates performance below the mean.

2. T-Score

Transforms $z$-scores into a standard scale with a fixed mean of 50 and a standard deviation of 10, eliminating negative numbers and decimals:

$$T = 50 + 10(z)$$

Worked Examples & LET Exam Scenario Traps

Worked Example 1: Identifying Skewness from Central Tendency

Data: In a 100-item Licensure Examination mock test, the statistical analysis revealed the following summary metrics: $\text{Mean} = 48$, $\text{Median} = 55$, $\text{Mode} = 62$.

Analysis:

1. Compare the three values: $62 > 55 > 48$, which means $\text{Mode} > \text{Median} > \text{Mean}$.
2. When the Mode is higher than the Median, and the Median is higher than the Mean, the tail extends to the left.
3. Conclusion: The distribution is Negatively Skewed, indicating that the mock test was relatively EASY and most examinees scored high.

Worked Example 2: Calculating Z-Scores and T-Scores

Data: A student obtained a raw score of $X = 85$ on a Professional Education test where the class $\text{Mean} = 70$ and $\text{Standard Deviation} = 10$.

Calculations:

1. Z-Score Calculation: $$z = \frac{85 - 70}{10} = \frac{15}{10} = +1.50$$ Interpretation: The student performed 1.5 standard deviations above the class average.
2. T-Score Calculation: $$T = 50 + 10(+1.50) = 50 + 15 = 65$$ Interpretation: The student's standardized T-score is 65.

LET Exam Scenario Trap: Percentile Rank vs. Percentage Correct

Trap: A examinee received a score report indicating she is at the 90th Percentile ($P_{90}$). A novice teacher tells her parents that she answered 90% of the test items correctly.

Correct Psychometric Interpretation: Percentile rank is a norm-referenced metric of relative position. Being at the 90th percentile means that the student scored equal to or higher than 90% of all examinees in the comparison group. It does NOT mean she got 90% of the questions right; she could have scored 55/100 on a very difficult exam and still landed at the 90th percentile if 90% of test-takers scored 54 or lower.