Factor Analysis (EFA vs CFA)

Question 1

What are items?

Answer 1

Questions in a questionnaire. Both the question and the person’s response are called items.
We use these to assess a psychological construct.

Question 2

What are scales/subscales?

Answer 2

These are the different behaviours/traits you are assessing within the questionnaire.
E.g. may have 6 items (questions), 3 looking at one particular scale and 3 looking at another particular scale.

Question 3

How do we compute a scale from items?

Answer 3

We find the mean scores/sum scores for the items that are measuring that particular scale.

Question 4

How do we compute a scale based on items?

Answer 4

We compute a mean score/sum score across all items. This should conceptually load on a specific scale.

Question 5

What does EFA mean?

Answer 5

Exploratory factor analysis.

Question 6

When is the only time EFA should be used?

Answer 6

When there is no prior assumption about which items load on which factors (constructs).

Question 7

What is the main goal of EFA?

Answer 7

To uncover the structure of how measured items load on unobserved constructs.

Question 8

What is the main issue with EFA?

Answer 8

When you develop a scale, you usually have some assumption about which items will load on which scales (subscales). This would not be exploratory.

Question 9

What happens if you run EFA and some items don’t load on the expected factor?

Answer 9

Use the new structure and re-label the factor.

This is really bad as it is a-theoretical.

Question 10

What does CFA stand for?

Answer 10

Confirmatory factor analysis.

Question 11

What is the main goal of CFA?

Answer 11

To test whether the data fits the hypothesised measurement model. To test whether the items load on the expected factors.
It is primarily theory driven.

Question 12

What do we do if we find that some items do not load on the right factors (CFA)?

Answer 12

We can drop the items and develop new items.

New scale development might take a lot of adjusting to find the best items.

Question 13

If each item has an error term, what would the true score of the reliability of each factor be (CFA)?

Answer 13

1.

Question 14

What type of modelling does CFA use?

Answer 14

Structural equation modelling.

Question 15

What are the two main steps to examining the factorial structure of a scale (CFA)?

Answer 15

1. Know how to set up the latent factors.
2. We set up our model in the way we assume the items will load on the latent factor - this allows us to compare if items load on sub-scales. Look at the model fit indices to see whether our model fits the data.

Question 16

Define validity.

Answer 16

If something measures what it is supposed to measure.

Question 17

Define reliability.

Answer 17

Consistency of a scale. When one person takes the same test several times and receives the same results.

Question 18

Why are latent constructs usually assessed with more than one question?

Answer 18

Latent constructs = psychological constructs that can’t be measured directly.
Often too complex to only be measured with one question. To assess reliability you must have multiple questions. Questions are slightly altered so they are measuring the same thing (consistency).

Question 19

What is the difference between scales and sub-scales?

Answer 19

Scales measure totally different things.

Sub-scales measure aspects that relate.

Question 20

Why is it sometimes better to calculate the mean score instead of the sum scores of items?

Answer 20

Because pps may not answer all questions.

Question 21

What is cross-loading?

Answer 21

Items from one sub-scale load on multiple items.

Question 22

Which is the stricter test - EFA or CFA?

Answer 22

CFA.

Theoretical assumptions from the start - test our scale from the beginning, therefore, we end up with a good scale.

Question 23

Why do latent factors represent the true score of a sub-scale?

Answer 23

Because only the common aspects across items load on a latent factor.

Question 24

What statistic do we use in SPSS to test items from a sub-scale for reliability?

Answer 24

Cronbach Alpha.