5 CFA – Analysis and Interpretation

Question 1

What is the difference between Exploratory and Confirmatory Factor Analysis?

Answer 1

Exploratory reduces a large number of variables to a smaller and related set. It investigates commonalities in items/variables to determine existence of ‘clusters’.

Confirmatory is used to seek evidence for a particular theory in a data set. On the basis of theory you can restrict the number of factors, the factor loadings (e.g. items that don’t define a given factor get 0 for that factor), the uniqueness of each item etc.

Question 2

What is RMSEA and how big do you want it to be?

Answer 2

The root-mean-square error of approximation. It measures goodness of model fit in the population. Should be less than .05. Or less than .1 at a pinch.

Question 3

What is the problem with using negatively worded items in factor analysis?

Answer 3

Sometimes they come out as defining their own factor, as people seem to respond to them in a systematic way.

Question 4

In missing data analysis, what is the difference between listwise, pairwise and EM (expectation-maximisation) methods?

Answer 4

• listwise - if a value on an item is missing, all items are ignored for that participant
• pairwise - if a value is missing, use all others
• Expectation Maximisation - if item is missing, EM imputes missing value based on how other people scored. For EM there must be less than 5% missing and it must be random.

Question 5

What is the MCAR assumption?

Answer 5

The assumption that data is Missing Completely at Random.

Question 6

In which situations might correlations be inflated or deflated?

Answer 6

Inflated –if similar items are used in scales measuring different variables (e.g. depression and SWL)

Deflated –if limited response range.