PsychometricProperties of
Standardized Tests
Types of Tests
Norm-Referenced Tests
Standardized
Comparison of individuals performance to the performance of a larger group (normative group)
“How does my client compare to the average?”
Normal distribution: Bell Curve
Dependent on the mean and the standard deviation
Uses measures of central tendency
Standard Deviation: determines the width, represents the distribution away from the average
Criterion-Referenced Tests
May or may not be standardized
Identify what a client can and cannot do compared to a predetermined criterion.
Assumes a level of performance that must be attained for a behavior to be acceptable
If the patient falls below that level, considered deviant
Used for: neurogenetic disorders, fluency disorders, and voice disorders.
Sometimes: aspects of articulation or language
Authentic Assessment Approaches
Alternative Assessment or Nontraditional assessment
Identifies what a client can and cannot do
Emphasizes contextualized testing
Test environment more realistic and natural
Ongoing: assessed during diagnostic and treatment phases of therapy
Client Portfolio: broad description of the client’s skills over the course of treatment and different settings
Client can practice self-monitoring and evaluation
Items included in portfolio: standardized tests, non-standardized tests, and treatment tasks
Requires more clinical skill, experience, and creativity
Assessed Qualitatively
Strategies:
Real-life simulations
Language sampling
Structured symbolic play
Videotaping
Audiotaping
Shipley, K. G., & McAfee, J. G. (2016)
Norm-Referenced
Normal Distribution
Reliability
Results are replicable- consistent results on multiple administrations
Types of Reliability
Test-retest- a tests stability over time
Split-half- tests internal consistency
Rater- level of agreement among individual rating the test
Two types:
Intrarater: if results are consistent when the same person rates the test on multiple occasions
Interrater: if results are consistent when more than one person rates the test
Alternate form or Parallel form- the test’s correlation coefficient with a similar test
Validity
The test measures what is claims to measure
Types of Validity
Face- test looks like it assesses the skill it assesses
Content- test’s contents are representative of the content domain of the skill being assessed
Construct- the test measures the theoretical construct it claims to measure
Criterion- test is related to an external criterion in a predictive or congruent way
Two types:
Concurrent: test compares to an established standard
Predictive: the test predicts performance in another situation or in the future
Shipley, K. G., & McAfee, J. G. (2016)
Sensitivity vs Specificity
Sensitivity: The rate at which a test can correctly identify a patient with a language impairment
Specificity: The rate at which a test can identify a child developing at a typical level of skills
Shipley, K. G., & McAfee, J. G. (2016)
Measures of Central Tendency
All considered an example of an average point, So is there a difference?
Mean: The mean takes all data and divides it by the amount of data contributed to get an average point (i.e. 2, 3, 4, 6, 6; mean is 4.2)
Median: Is the middle number of all data points when lined up in order from smallest to largest. (i.e. 2, 3, 4, 6, 6; the median is 4)
Mode: Is the most frequent data found in a data set. Useful with a numeric or non-numeric representation (i.e. 3, 4, 6, 6, 2; the mode is 6 or blue, blue, orange; the mode is blue).
Note: In normal distribution the mean, median and mode are the same as there is an equal amount of values or data on both sides of the middle point.
Confidence Intervals
Tests scores are estimates. How do you know if the score is a correct representation of your client?
Confidence intervals is the degree of certainty on the part of the test developers that the statistical values are true.
Takes into account natural human variability.
Most tests provide statistcal data for a confidence interval of 90% - 95%. (textbook)
The higher the confidence interval the more reliabile the test.
Provide the clinician a range of reliable scores, increasing the probablity of the true score
Shipley, K. G., & McAfee, J. G. (2016)
Basals and Ceilings
Allows the examiner to make use of the most relevant testing materials to the specific examinee and reduces the items being administered. Based on client's age. Clinican must review standardized test manuals to exam the predetermined basals and ceilings starting points and ending points. Not all standardized test presents with Basals and Ceilings rules. (i.e. the GFTA-3)
Shipley, K. G., & McAfee, J. G. (2016)
Basal
Starting point for test administration and scoring.
Determined by the number of consecutive correct responses.
Ceiling
Ending point for test administration and scoring.
Determined by the number of consecutive incorrect responses.
Shipley, K. G., & McAfee, J. G. (2016)
Standarized Test Evaluations/
Diagnositics Inventories
https://drive.google.com/drive/u/1/folders/1F6_8rTrFkZ391JcT5JGAS6Fe3L1nMSdl
Link above includes the following assessments:
BDAE- Boston Diagnostics Aphasia Examination
CELF-5 - Clinical Evaluations of Language Fundamentals
CLQT- Cognitive Linguistic Quick test
GFTA-3 and KLPA-3- Goldman-Fristoe Test of Articulation and Khan-Lewis Phonological Analysis
PLS-5 - Preschool Language Scales
Standardized test assessment:
https://nyu.box.com/s/eqh2e6wz25geahd589qkc818vdv5p6bk
NYU’s Speech and Hearing clinic
References
Lissa Power-deFur Sensitivity versus Specificity, August 8, 2016
https://www.speechpathology.com/ask-the-experts/sensitivity-versus-specificity-3334
Connecticut State federal laws for assessment and diagnosis
https://leader.pubs.asha.org/doi/10.1044/leader.PA.23042018.28
Shipley, K. G., & McAfee, J. G. (2016). Assessment in speech-language pathology: A resource manual (6th ed.). Boston, MA: Cengage.
Mean, Median, Mode review. Khan Academy. retrieved on September 12, 2021. https://www.khanacademy.org/math/statistics-probability/summarizing-quantitative-data/mean-median-basics/a/mean-median-and-mode-review