Percentile Scores, Reference Groups, and Score Interpretation
This paper examines the use of percentile scores as descriptive statistics in psychological testing and assessment. It discusses how percentile ranks represent a person's standing relative to others on a given measure and identifies key drawbacks in their interpretation. Specifically, the paper addresses the problems that arise when skewed score distributions are treated as normal distributions, and when percentile scores are interpreted without reference to an appropriate comparison group. Using examples from mathematics aptitude testing, the paper illustrates how the choice of reference group dramatically affects the meaning of a percentile rank. It concludes by emphasizing the importance of matching percentile scores to appropriate norms and distributions.
- Percentile Scores as Descriptive Statistics: Defining percentile ranks and their purpose in assessment
- The Problem of Skewed Distributions: Why skewed data distorts percentile rank accuracy
- The Role of the Reference Group: How reference group choice determines percentile meaning
- Separate Norms and Appropriate Comparisons: Using proper norms to overcome interpretive drawbacks
✍️ How to write this paper — guide, tools & examples ▾
What makes this paper effective
- Uses a concrete, recurring example (mathematics aptitude and fourth-graders vs. PhD holders) to make an abstract statistical concept immediately accessible.
- Moves logically from defining percentile scores, to identifying distributional problems, to identifying reference-group problems — each issue building on the last.
- Grounds every claim in multiple cited sources, demonstrating appropriate academic support for a short expository piece.
Key academic technique demonstrated
The paper demonstrates concept illustration through counterexample: rather than simply defining what a percentile rank is, the author shows exactly how it can mislead — first through distributional distortion (skewed data treated as normal) and then through inappropriate reference groups. This technique is effective in methodology-focused writing because it makes the stakes of misinterpretation concrete and memorable.
Structure breakdown
The paper opens by defining percentile scores within the broader context of descriptive statistics, then devotes a paragraph to the distributional threat (skewness), a full paragraph to the reference-group threat with an extended example, and a final paragraph that synthesizes both problems and points toward best practice. The reference list follows APA format. The structure is tight and thesis-driven for its length.
Percentile Scores as Descriptive Statistics
A person's ability, skill, or standing relative to other people is often assessed in order to design an appropriate intervention (Cohen & Swerdlik, 2013; Runyon, Coleman, & Pittenger, 2000). Summary statistics (descriptive statistics) allow an assessor or researcher to summarize raw data and scores from individuals being tested. The percentile rank, or percentile score, represents a person's standing relative to other individuals on a particular test (Cohen & Swerdlik, 2013).
The Problem of Skewed Distributions
There are notable drawbacks to percentile scores. If a percentile rank is obtained without careful consideration of the underlying data, it may not be a meaningful or accurate summary statistic. For example, if one uses a standard distribution such as a Z-distribution to determine the percentile rank of a particular score, but the distribution of raw scores is not normal, then the percentile rank will not accurately reflect a person's standing on that measure (Huck, 2012). Perhaps one of the biggest mistakes made in assessment and testing is treating highly skewed distributions of scores as though they were normal distributions (Cohen & Swerdlik, 2013). This error occurs frequently with physiological measures such as reaction time, which typically follows a positively skewed distribution — the majority of people score at the lower end of the distribution, with fewer people scoring at the higher end (Runyon et al., 2000). Transforming a highly skewed distribution into a Z-distribution results in an inaccurate representation of the scores (Cohen & Swerdlik, 2013).
References
Cohen, J. R., & Swerdlik, M. (2013). Psychological testing and assessment: An introduction to tests and measurements (8th ed.). New York: McGraw-Hill Education.
Huck, S. W. (2012). Reading statistics and research (6th ed.). Columbus, OH: Allyn & Bacon.
Runyon, R. P., Coleman, K. A., & Pittenger, D. J. (2000). Fundamentals of behavioral statistics (9th ed.). New York: McGraw-Hill.
Create your account
Always verify citation format against your institution’s current style guide requirements.