Major Milestones In Testing

Front 1

Edouard Seguin (1812-1880)

Back 1

MR; Form board for discrimination & motor control.

Front 2

Alfred Binet (1857-1911)

Back 2

Differentiate students for instruction (1896; 1905).

Front 3

Sir Francis Galton (1883)

Back 3

Individual differences research & measurement

Front 4

James Cattell (1890)

Back 4

Mental Test.

Front 5

1900-20: Group achievement tests;

Back 5

Spearman’s testing theory; Army Alpha and Beta.

Front 6

20-50’s: Psych Corp born; mental measure yearbooks;

Back 6

Wechsler-Bellevue; Gesell Maturity Scales; MMPI appeared. First standards.

Front 7

Twentieth Century:

Back 7

-Strong demand
–High optimism
–Century of extensive development

Front 8

Today…Tomorrow?

Back 8

•Computer Assisted
•Increased Design Complexity
•Easy Methods
–Administration
–Application
–Scoring
•Web based assessment

Front 9

Basic Psychometric Principles
•Tests translate observations of behavior into numbers.
–Behavior and/or attitudes into quantities.

Back 9

•Measurement involves the establishment of rules for assigning numbers to objects.
•To make sense of measurement, scales are needed to place numbers in relationships.

Front 10

Common Scale Types
–Nominal: name only.
–Ordinal: rank order, has magnitude, does not have a zero point or equal distance between points.

Back 10

–Interval: has magnitude, equal intervals, no zero.
–Ratio: has magnitude, equal intervals and an absolute zero.

Front 11

•Norm-referenced measurement

Back 11

–Performance compared to specific group.

Front 12

Descriptive statistics
•Summarize data.
•Allows comparison.
•Central tendency- mean, median, mode.

Back 12

•Dispersion
–Variance: measure of variability of a group of scores
–Standard deviation-square root of variance.
–(SD) Measure of the extent to which scores deviate from the mean.

Front 13

Nominal:

Back 13

name only.

Front 14

Ordinal:

Back 14

rank order, has magnitude, does not have a zero point or equal distance between points.

Front 15

Interval:

Back 15

has magnitude, equal intervals, no zero.

Front 16

Ratio:

Back 16

has magnitude, equal intervals and an absolute zero.

Front 17

Score Distributions
•Normal curve is most commonly used.
–Tells us exactly how many cases fall between any two points on the curve.

Back 17

–Allows us to use standard scores (transformed to a given mean and standard deviation).
–The use of the normal distribution curve allows comparisons and establishes relative position.

Front 18

Standard Scores
•Defines the percent of cases that fall under the normal curve.
•A basic Z score has a mean of 0 and SD of 1.

Back 18

•Deviation IQ
•McCall’s T scores
•Stanine scores (Standard Nine)

Front 19

Z Score Example
A ‘z’ score is the obtained score minus the test mean divided by test’s standard deviation.
•The mean on a test was 100; standard deviation 20; your score 125.
•The z score is (125-100)/20=25/20 or 1.25
•Express this score as a T score.
•z =1.25=(X-50)/10; 12.50=x-50; 62.50
•T=10Z + 50
•What would the T score be if the z were negative?
•T=10(-1)+50=40

Back 19

Z Score Comparisons
•A class test has a SD of 5 and a mean of 20; Your score was 15.
•What would be your T score?
•If this test were the GRE, what would be your score?
•Approximately what percentage of the students taking this exam performed better than you?
Steps:
 First, determine the z score.
 Second, fill in the formula with fixed variables.
 Third, do the calculation to complete the transformations of the score.

Front 20

Variance
•Issue: the extent to which differences in scores are attributable to “true” differences in the characteristics under consideration and the extent to which they are attributable to chance errors.

Back 20

•Error or random variance is any condition that is irrelevant to the purpose of the test.
•Chance fluctuations contributes to error.
•‘True’ variance is about relevant differences.
•Standardized administration is intended to reduce error variance.

Front 21

Establishing Reliability
•Correlations are used to look at the relationship between sets of scores.
•Reliability coefficients usually fall in the .70s, .80s or .90’s.

Back 21

•The more important the prediction or application, the more critical the reliability.

Front 22

Standard Error of Measurement
•Psychologists assume some test “unreliability”.
•Observed Score=True Score + E (error)
•Measurement error is assumed to be random.
–The errors would form a normal distribution.

Back 22

•Standard deviation of errors is the SEM or the standard error of a score.
•Provides information regarding confidence in the obtained test score.
•An obtained score is not assumed to be the ‘True’ score.
•The larger the SEM, the less confidence there is that the obtained score is close to True score.

Front 23

Confidence Intervals
•A Confidence interval is a band of scores around the obtained score in which we assume the true score is located.
•Intervals are set depending upon the degree of confidence required. The greater the need to hit the true score, the greater the confidence interval.

Back 23

•Use Z scores to create 68%, 95%, or 99% CI. Convention is typically 95% level.
•Most tests provide tables for Confidence Intervals.
•No confidence interval can be constructed that can predict with absolute certainty where a person’s true score will lie.

Front 24

Reliability & Validity

Back 24

•A test must first be dependable (Reliability).
•Once a test is dependable, the ongoing process of validation can begin (Validity).

Front 25

Validity

Back 25

The extent to which a test measures what it is supposed to measure and the appropriateness with which inferences can be made on the basis of the test results.

Front 26

Reliability

Back 26

Consistency of scores obtained by the same person when re-examined with the same test on different occasions, or with different sets of equivalent items, or under other variable examining conditions.

Front 27

Establishing Reliability
•Test-retest method
–The error variance corresponds to the random fluctuations of performance from one test session to another.

Back 27

–Interval of time is always specified.
–The higher the retest reliability, the more consistent the tests scores are from administration to administration.

Front 28

Test-Re-test Drawbacks
•Practice and Carry over effects
–Carry over is associated with the content of questions.

Back 28

–Practice effects are associated with the type of questions.
•Due to these effects, test-retest reliability is not always an appropriate measure of reliability.

Front 29

Alternate-Form Reliability
•Two tests are created. Scores are correlated as an estimate of reliability.

Back 29

•The error variance may reflect problems with content sampling.
–A test with strong reliability covers domain fairly.
–This method does not account for errors due to inappropriate item sampling.
•Alternate-Form Reliability does correct some for carry over effects but not fully for practice effects.

Front 30

Reliability Coefficients
•Split-half reliability (cut test in half)

Back 30

–Consistency of content sampling is an issue.
–Longer tests are more reliable.
–Spearman Brown formula can provide correction.

Front 31

•Kuder-Richardson & Coefficient Alpha
–Error variance reflects content sampling and heterogeneity of behavior sampled.

Back 31

–KD is mean of all split-half coefficients; used on True/False or right/wrong items.
–Coefficient Alpha is used on numerical scores.

Front 32

Scorer Reliability
•This is error due to rater differences.
•Tests can be complicated to score.

Back 32

•Tables may be confusing to read.
•Tests designed for ease of administration and scoring increase scorer reliability
•Computer scoring programs eliminate this factor.

Front 33

Recommendations for Practice
•Become familiar with the tests used in your area of practice. Know about reliability and validity.

Back 33

•Do read the manual on the tests in use in your practice area, even if you do not directly administer.
•Understand the meaning of any scores that are reported to the clients in your practice area.