Confirmatory factor analysis

8 important questions on Confirmatory factor analysis

What is the main difference between EFA en CFA regarding correlations between factors and variables?

All observed variables are allowed to correlate with each factor --> EFA.

The factors and correlations among factors and variables are determined a priori --> CFA.

What are the main similarities and differences of EFA and CFA?

Similarities:
- they both try to determine the number and nature of latend variables or factors that account for variation among a set of observed variables. (reflective)


- Both try to reproduce the relationship among a set of indicators with smaller set of latend variables.


Differences:
EFA
- Data driven
- Standardized observed variables
- Correlation matrix is analyzed
- Errors are assumed to be uncorrelated

CFA
- Theory driven
- Unstandardized observed variables
- Evaluation based on how the sample covariance is reproduced
- Error may be correlated.

What is evaluation of measurement invariance and with which of the two types of factor analysis can you check this?

Measurement invariance is that e.g. Different cultures tend to answer the questions differently. E.g. German people do not like to answer extremes. You can check this with Confirmatory factor analysis.
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What do you do in the first step of CFA, model specification?

In this step you determine which constructs you enter to the model, including determining the relationship between them and the measurement model of each construct. Each construct (or factor, or latent variable) is identified priori; fixed, free or constrained.

Within the step model specification, the number of factors and observed variables that load on a construct are specified in advance. What is the difference between setting values fixed, free or constrained?

Fixed - values of fixed parameters are set by the researcher.
Free - parameters will be estimated
Constrained - parameters wil lie estimated acknowledging the constraints

What do you determine in the second step of CFA: Identification? What is the difference between the deductive and inductive part?

Identification is basically the research design.

Deductive: the model structure and parameter values (fixed, free or constraint) determine the variances and covariances of the observed variables.

Inductive:comparing the empirical values with the model. How close are these to the original values? You want the differences to be as small as possible. The empirical variances and covariances yield estimates of unknown (free) parameter values given the structure of the model.

What do you do within the third step of CFA: estimation?

Estimation includes that you try to estimate the parameters in such a way the difference between the sample covariance matrix and the implied covariance matrix is minimal.

Estimation techniques:
- Maximum likelihood (ML) = most common approach. Maximizes the likelihood of getting a model with minimum discrepancy between the sample covariance and implied covariance matrix.
- Unweighted least squares (ULS)
- Generalized least squares (GLS)
- Distribution-free (ADF)

In the fifth step, eventually specification, you can improve the model fit. How can this be done and what is the danger you should avoid?

By means of respecification: adding or removing items. Avoid overfitting --> don't get too guided by model fit, because you do confirmatory FA and not exploratory!

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo