Test score

A test score is a piece of information, usually a number, that conveys the performance of an examinee on a test. One formal definition is that it is "a summary of the evidence contained in an examinee's responses to the items of a test that are related to the construct or constructs being measured." [Thissen, D., & Wainer, H. (2001). Test Scoring. Mahwah, NJ: Erlbaum. "Page 1, sentence 1."]

Test scores are interpreted with a norm-referenced or criterion-referenced interpretation, or occasionally both. A norm-referenced interpretation means that the score conveys meaning about the examinee with regards to their standing among other examinees. A criterion-referenced interpretation means that the score conveys information about the examinee with regards a specific subject matter, regardless of other examinees' scores. [Iowa Testing Programs guide for interpreting test scores [http://www.education.uiowa.edu/itp/itbs/itbs_interp_score.htm] ]

Types of test scores

There are two types of test scores: raw scores and scaled scores. A raw score is a score without any sort of adjustment or transformation, such as the simple number of questions answered correctly. A scaled score is the results of some transformation applied to the raw score.

The purpose of scaled scores is to report scores for all examinees on a consistent scale. Suppose that a test has two forms, and one is more difficult than the other. It has been determined by equating that a score of 65% on form 1 is equivalent to a score of 68% on form 2. Scores on both forms can be converted to a scale so that these two equivalent scores have the same reported scores. For example, they could both be a score of 350 on a scale of 100 to 500.

Two well-known tests in the United States that have scaled scores are the ACT and the SAT. The ACT's scale ranges from 0 to 36 and the SAT's from 200 to 800 (per section). Ostensibly, these two scales were selected to represent a mean and standard deviation of 18 and 6 (ACT), and 500 and 100. The upper and lower bounds were selected because an interval of plus or minus three standard deviations contains more than 99% of a population. Scores outside that range are difficult to measure, and return little practical value.

Note that scaling does not affect the psychometric properties of a test, it is something that occurs after the assessment process (and equating, if present) is completed. Therefore, it is not a psychometric issue, but a public relations issue.

References