Psychometrics - Alpha Cont.

Question 1

What is co-variance?

Answer 1

A measure of how much score on items go together - if a person who does well on item 1 does well on item 2 then item 1 and 2 have a lot of co-variance

Question 2

Why do we need to be careful when saying that alpha is the mean of all split-half reliabilities?

Answer 2

Split-half relies on inter-item covariance, which alpha measures. But it also measures variance, which isn’t a part of correlation. Variance introduces error, which lessens the value. So alpha is actually smaller than the mean of all possible split half reliabilities.

Question 3

When does alpha not underestimate the mean of all split-half possibilities?

Answer 3

1. When all the standard deviations on the test items are equal
2. When you use Flanagan-Rulon correction for half length and not SB bc FR accounts for variance

Question 4

What does first factor saturation mean?

Answer 4

The extent to a test is made up of one factor, in the sense that it is unidimensional

Question 5

Why is it wrong to think that a higher alpha means that there is only one factor in the test?

Answer 5

Because internal consistency and homogeneity are not the same thing

Question 6

What is the difference between internal consistency and homogeneity?

Answer 6

Internal consistency: how inter-related the items are
Homogeneity: unidimensional
A test can be internally consistent and still measures multiple factors, if these factors are related to one another

Question 7

When do we use the standard error of measurement?

Answer 7

When we went to calculate the range in which a person’s true score should fall. We create confidence intervals once we’ve found the SEM

Question 8

Formula for SEM

Answer 8

SEM = Stddev*SQRT(1-alpha)