Most frequently occurring number in the distribution
o
Least common measure of central tendency
o
Variability
o
Dispersion of numerical values
o
Range
Subtract lowest score from highest score
o
Standard deviation
Spread of scores around the mean
o
Correlation
o
The correlation of the coefficient reflects the degree of linear relationship between two
variables
o
Range from -1.00 to +1.00
o
Tell direction and magnitude
o
Direction
Positive (+)
Negative (-)
o
Magnitude
Tells strength of the relationship
o
Regression

o
Regression analysis is a statistical technique used to predict criterion performance based on
predictor scores
o
Y=a+bX
Y: predicted criterion score
a: intercept
b: slope
X: predictor score for an individual
Lecture 3
Reliability
o
Predictor: any variable used to forecast a criterion
o
Judge the goodness of a predictor with two psychometric criteria
o
Reliability
Consistency, stability, or equivalence of a measure
Four types
Test-retest
Equivalent form
Internal consistency
Inter-rater
o
Validity
Test-retest
o
Reveals the stability of test scores upon repeated applications of the test
o
Generally, shorter time interval = higher test-rest reliability
Equivalent forms reliability
o
Reveals the equivalence of test scores between two versions of the same test
o
Difficult to construct two measures with similar meanings and properties
Internal consistency reliability
o
Reveals homogeneity of items comprising a test
o
Homogenous = item content is similar
o
Generally, the longer the test, the greater its reliability
Inter-rater reliability
o
Reveals the degree of agreement among the assessments of two or more raters
o
Establishes whether behavior was reliably observed
Reliability vs. Validity
o
Reliability = stability, consistency
o
Validity = accuracy, correctness
o
Accuracy or appropriateness of drawing inferences from test scores
o
Reliability is inherent in a measuring device; validity depends on the use of the test
Construct validity
o
Construct = abstract, theoretical concept proposed to explain aspects of behavior
o
E.g. intelligence, leadership, motivation

o
Construct validity = degree to which a test is an accurate measure of the construct it purports to
measure
o
There should be evidence of
o
Convergent validity
Scores on a test converge with scores on other known tests of the construct
o
Discriminant validity
Scores on a test are unrelated to scores on tests of other constructs
Criterion-related validity
o
Degree to which a test forecasts or is statistically related to a criterion
o
Validity coefficient = correlation of predictor scores with criterion scores
o
Desired range of validity coefficients is about 0.30-0.40


You've reached the end of your free preview.
Want to read all 22 pages?
- Fall '14
- SarahJaneRoss
- Psychology, researcher