Test construction Flashcards
(17 cards)
Incremental validity
Criterion validity; amount of improvement in success rate from using a predictor test
Is optimized when base rate is moderate and selection ratio is low
Concurrent Validity
Criterion validity; predictor and criterion are measured close in time
Predictive validity
Criterion validity; Significant gap in time between measuring the predictor and the criterion.
Face Validity
Level of test subjectively appears to measure what it’s intended to
Consequential validity
Addresses whether a test has a positive or negative social consequence.
Convergent validity
Construct validity; correlates a test with other available measures of the same trait
Monotrait-heteromethod; high correlations establish convergent validity
Divergent (discriminant) validity
Construct validity; correlates a test with measures of different traits. Should have low correlation with other measures of different constructs
Heterotrait- monomethod; low correlations establish divergent validity
Sensitivity
Tests ability to accurately identify true positives
TP divided by TP + FN
Specificity
Tests ability to accurately identify true negatives
TN divided by TN + FP
3 types of validity and their subtypes
Content, Criterion; concurrent, predictive, Construct; convergent, divergent
Reliability coefficient
Rxx; ranges from 0.0 (no reliability) to 1.0 (perfect).
Correlation between two sets of scores x and x.
Minimum acceptable value is 0.8
Never squared; if rxx is 0.85 the true score variability would be 85%
Test-Retest
Determines tests consistency over time.
People administered the same test at two different times.
Reliability decreases the longer the time between tests.
Alternate forms
Also known as: parallel forms or equivalent forms.
Similar, but not identical versions of a test are administered to the same group at two different times
Z-score
Mean of 0 standard deviation of 1.0
T-score
Mean of 50 and SD of 10
Stanines
Mean of 5 and standard deviation of 2 ranges from 1-9.
Conversion of raw scores to percentile ranks
Non linear transformation since percentile rank distribution is always rectangular (flat) regardless of the shape of the distribution raw scores.