Reliability

What is classical test theory?

Statement 1: true score theory is the idea that every measurement is split into two parts (actual and error) Statement 2: classical test theory is the concept that every measure is split into two parts (observed and actual)

Statement 1: when calculating the reliability use variance because it is additive Statement 2: reliability in terms of the relationship between true and total variance is true variance/total variance

Which one is NOT a source of measurement error?

Statement 1: we can always estimate reliability directly using true variance and measurement error Statement 2: we can measure reliability using four different types of tests: inter-rater, test-retest, alternate forms and external consistency

What is used to calculate internal consistency when there are more than two possible outcomes?

Statement 1: KR-20 is used instead of Cronbach's alpha when there is a question with only two outcomes Statement 2: Cronbach's alpha is calculated by splitting the data in half, calculating total scores and correlations, repeating for all possible two way splits

Statement 1: Parallel forms are where the two versions of a test have equivalent content and difficulty level Statement 2: Alternate forms are tests where the two versions of the test have a shared mean, SD and correlations with other tests

What measure shouldn't you use when there is a heterogeneous test?

Statement 1: when there is a static test it is possible to use test-retest but not when the test has dynamic characteristics Statement 2: when there is a restriction of range reliability estimates are still able to be interpreted

What estimate of reliability would not be appropriate for speed tests?

Statement 1: increasing the number of items in a test continues to double the amount of reliability Statement 2: the spearman brown formula is used for estimating reliability when a test is shortened or lengthened

What is the standard error of measurement?

How many scores will be within 1 SEM of the true score (+/- 1 SD)?

How many scores will be within 1 SEM of the true score (+/- 1 SD)?

Statement 1: standard error of the difference is used to work out the margin of error in an individual's score Statement 2: in order for two scores to be significantly different they have to differ by at least 2 standard error of the differences (1.96)

What is the reliable change index? (RCI)