Exam 1 - Assessment (Psychometrics)
standardized tests
aka formal tests have standard procedures for administration and scoring most, but not all, are norm-referenced
types of reliability
test-rest, split-half, rater, alternate form
normal distribution
"Provides a range of scores by which others are judged when they take the same test." Depicted with a Bell-Shaped curve Symmetrical Height and width are dependent on two quantities • Height = Mean • Width = Standard Deviation
psychometrics
"The measurement of human traits, abilities and certain processes."
common errors (norm-reference tests)
- Measuring treatment progress - STATISTICAL REGRESSION: phenomenon states that subjects earning extreme scores (high or low) tend to regress to the mean in subsequent administrations - Analyzing individual test items for treatment target selection - Forgetting that formal tests almost always distort what they are designed to examine
raw score
- Number you arrive at when "scoring" the exam - Often represents the number of correct items on the test - e.g, If 75 items were administered and client responded correctly to 50 items, the RAW score would be 50 - Not meaningful by themselves
percentile rank
- Reflects the percentage of subjects or scores that fall at or below a particular score - 50th percentile represents the median (middle) performance of the normative sample - Consistent performance below 10th percentile is cause for clinical concern
adjusted age
- Takes into account the gestational development missed due to premature delivery - Determined by using the child's DUE DATE, rather than birth date - Becomes less relevant as the child grows and considered irrelevant after the age of 3
reliability
- Test's ability to yield replicable results even when time, examiner, and/or other variables change. - When administered properly, the test should yield consistent results on repeated administrations. - Is the test CONSISTENTLY measuring what you want?
validity
- The degree to which a procedure actually measures what it purports to measure. - Does the test measure what it says it measures? Related to the PURPOSE for which the test is used - EXAMPLE: If you are testing language comprehension, you need to select a test that measures that.
ceiling
- end point for a test - TYPICALLY assumed the client would incorrectly answer the remaining items
basal
- starting point for a test - TYPICALLY assumed the client would answer all previous items correctly
empirical rule
68% of all outcomes will fall within one standard deviation of the mean 95% of all outcomes will fall within two standard deviations of the mean 99.7% will fall within three standard deviations of the mean
modifications
Changes to the standardized administration protocol
advantages of norm-reference tests
Comparable Efficient Recognizable Easy to Administer Preferred by insurance companies and schools
age/grade equivalents
Depicts the average performance for a particular age-group LEAST useful scores to be obtained from standardized exams
face validity
Does is look like what it purports to measure?
construct validity
Does the test appropriately measure a theoretical construct?
content validity
Does the test take all of the content into consideration?
confidence interval
Estimation of the likelihood a score would fall within a range of scores centered around the true score Usually expressed as: 68%, 90%, 95% Important component to consider when making a diagnosis
chronological age
Exact age in years, months, days
criterion-related validity
How well does the test relate to other tests that measure the same skill/behavior?
rater reliability
Level of agreement among individuals administering/scoring the test
central tendency
Mean (Median, Mode) Arithmetic average
standard/scaled scores
Mean of 100, standard deviation of 15 (i.e., Average Range: 85-115) Mean of 10, standard deviation of 3 (i.e., Average Range: 7-13)
accommodations
Minor adjustments to a testing situation that DO NOT compromise standardization
criteria for disorder
Paul: Score <10th %ile (-1.25 SD, standard score of 80) on TWO well-constructed measures AND Perceived as having a problem by parents, teachers, or other communication partners Text book: Performance near bottom 5% of the normal distribution 1.5 - 2.0 standard deviations below the mean In other words: compare results with other information collected... NORM-REFERENCED tests are just ONE piece of the puzzle!
split-half reliability
Test's internal consistency
test-rest reliability
Test's stability over time
variation
Standard Deviation Distribution away from group average
disadvantages of norm-reference tests
Static Inflexible administration Not representative of "real life" Only evaluate isolated skills Invalid for culturally and linguistically diverse populations
alternate form reliability
Two forms of the same test are given
types of validity
face, content, construct, criterion-related
criterion reference test
• Performance is determined in relation to a particular standard • Standardized OR Nonstandardized • Commercial (e.g., RITLS, IRI, FCP) • Clinician Created (e.g., rating scales, accuracy measures, language samples)
norm-reference test
• Results compared to performance of a normative sample • Standardized • Commercial (e.g., OWLS-II, CELF- 5, PLS-5, etc.)