4 Path Analysis

Question 1

What does a goodness-of-fit test tell us?

Answer 1

How well expected correlations match the observed correlations

Question 2

What is the most commonly used estimation method of goodness-of-fit?

Answer 2

Maximum likelihood (ML) Estimation chi-square statistic

Question 3

What is a limitation of chi-square test?

Answer 3

Quite sensitive, particularly with large sample sizes

Question 4

What do absolute fit indices tell us?

Answer 4

How well the implied model fits the sample data

Question 5

How are incremental fit indices calculated?

Answer 5

Compared with a null model (where variables are specified not to correlate) and whether specified model is incrementally better compared with the no-relationship model

Question 6

What are the two absolute fit indices and their recommended cut-offs?

Answer 6

RMSEA (.08 or less, although .11 for small samples sufficient) and SRMR (.06 or less, .08 for small samples sufficient)

Question 7

What are the two incremental fit indices and their recommended cut-offs?

Answer 7

CFI and TLI (above .90 satisfactory, above .95 regarded as good fitting)

Question 8

What is the MI statistic?

Answer 8

A chi-square value representing an approximate reduction in the Chi-square ML ( > 3.84 for each df reflects a significant improvement)

Question 9

What is the EPC?

Answer 9

Expected parameter change, representing the EPC if it were to be estimated in the model.

Question 10

What are we looking for in the standardised residual covariance matrix?

Answer 10

95% of values to fall between -2 and +2 (otherwise, a problematic level of error existing in the model).