Reliability Flashcards
Fall 2018 Midterm
What are observed scores (according to the true and error scores theory of reliability)?
Observed score (X) is made up of two independent components: latent true score (t) and latent error score (e).
What are errors (according to the true and error scores theory of reliability)?
Errors are accidental deviations that are different in every individual case, occur impartially in every direction, and may randomly augment or diminish observed values.
What are the 5 assumptions of the true and error scores theory of reliability?
1) True scores reflect stable characteristics through time.
2) Observed score includes random error. X = T + E
3) Mean error is 0.
4) Covariance of true score with error score is 0.
Covariance of observed score = variance of true score + variance of error score.
5) Correlation of observed score with true score = variance of true score/variance of observed score.
Correlation between T + E = 0, correlation between two error scores is 0.
What are the 4 assumptions of the parallel tests theory of reliability?
1) There exists parallel tests for any variable.
2) Tests have equal variance.
3) All correlations among parallel tests are equal.
4) Trues scores are just a person’s average of those parallel tests.
The theory of parallel tests is based on what kind of reliability?
Alternate-form reliability
The theory of true and error scores is based on what kind of reliability?
Test-retest reliability
What is test retest reliability?
If you give the same test to the same people twice their scores should be the same.
- Time interval needs to be shorter than actual change that might occur in the attribute
- Interval must be short enough to preclude real change in the true scores (for coefficients over any interval)
What is internal consistency?
Whether several items proposed to measure the same construct produce similar scores.
Reliability estimate based on number of items and average intercorrelation among the items. Looks at composite rather than individual items.
What is the Spearman Brown formula used for?
Measures reliability as a function of the number of items and the average intercorrelation among them (higher correlation/higher sample size -> higher reliability).
Assumes equal variances.
What is coefficient alpha?
Measures reliability as a function of the number of items and the intercorrelation EXCEPT replace average correlation in SB with average variance among items.
Does not assume equal variances.
What is the correction for attenuation?
Measurement error attenuates a correlation, apply correction to see what measure would be with perfect reliability. ‘Rids a correlation from the weakening effect of measurement error’
Correlation is divided by the geometric mean of the reliability coefficients of two tests.
What is the standard error of measurement?
Distribution of observed scores around the true scores if we measure something multiple times. Used to estimate a person’s true score given their observed score, and be used to generate confidence intervals around the estimated true score.
What does reliability ask?
Would you get the same results from a different instrument?
