What is the difference between first and second order Confirmatory factor analysis "CFA" ?
Hi Nader,
the structure of both types is exactly the same--namely the "common factor model" which claims that a latent variable is causally influencing lower level variables and, hence, is responsible for the correlations among the lower level variables.
In the case of a first order CFA, the lower level variables are error-containing measured/observed variables. These can be observations in a field study or responses in a questionnaire survey.
In the case of a second order CFA, the lower level variables are themselves latent variables (which themselves have effects on their measured indicators. A famous example is the g-factor that is responsible for differences in domain-specific forms of intelligence (e.g., numeric vs. verbal intelligence). These specific variables themselves may be latent variables having effects of actual tests (e.g., solving numerical tasks).
Please note that in both cases the latent variables do NOT differ with regard to their dimensionality. I say this because often, people rather think of a second order latent variable as the sum of the primary latents which is wrong. The g-factor is as singular and narrow as each of his domain-specific sub-factors. Note: Factor = variable = dimension. More dimensions = different variables.
HTH
Holger
I'm with Holger's opinion. In addition, you may also read publications applying second-order latent variables:
Article Methods for modelling reflective-formative second order cons...
Regards
1st order models only have item-construct loadings.
2nd order models have item-1st construct loadings + 1st construct-2nd construct loadings.
Hope this helps,
Matt
Dear Nader Mohamed
It is important for the operational definitions and measurement Instrument of the study variables. They are used in latent constructs.
Let me explain with an example. Suppose that you are studying the impact of burnout on health quality of life of the participants by using SEM for CFA. In this case, burnout is one of your variables. Burnout has three sub-dimensions: personal burnout, work related burnout, and client oriented burnout. Here, burnout is your second order factor endogenous variable, while three sub-dimensions are your first order factors.
Please check the following article. There are some first order and second order factors discussed in this study.
Thesis THE IMPACT OF MOBBING AND JOB-RELATED STRESS ON BURNOUT AND ...
simply the second order is the indicator plus items and the first order is the main variable.
In the case of a first order CFA, the lower level variables are error-containing measured/observed variables. ... In the case of a second order CFA, the lower level variables are themselves latent variables (which themselves have effects on their measured indicators.
For first order CFA, we check only discriminant validity and for second order CFA, we check convergent validity. If the constructs have multicollinearity, only then second order of CFA is required otherwise not.
- Badges
- Science topic
- Similar topics
- Correction