first order and second order confirmatory factor analyses are conducted during scale validation of multidimensional constructs. we need to design both types of models and which ever has better results in terms of beta coefficients and goodness of model fit is accepted.

first order and second order confirmatory factor analyses are conducted during scale validation of multidimensional constructs. we need to design both types of models and which ever has better results in terms of beta coefficients and goodness of model fit is accepted.

Hello Hamid,

I would like to help you, but I need to be clearer about your questions / issues first. Could you let me know how many items you fed into your initial factor analysis please?  How many "respondents" do you have?  Did you get satisfactory results for your KMO index and Bartlett's test of sphericity? Did you ask for a scree plot, and, if so, how many factors seemed to be in your data? Did you restrict your subsequent analysis to that number of factors? What kind of analysis did you ask for:  - for example, an oblique or an orthogonal rotation?  Did you remove any items as a result of your first factor analytic results?

I hope I am asking the right questions and that you understand what I mean.

Would you be happy to attach your scree plot and the results of your initial rotated solution? (Are you using SPSS?  If so, it is easy to transfer its output to a Word document, and then that could be attached. Please let me know if you don't know how to transfer SPSS output to a Word document if you'd like advice re that. In essence, you just click your cursor in the middle of the appropriate SPSS output, copy it, and then paste it into a blank Word document.)

I also don't understand what you mean by "convergent validity is not fulfill". Could you expand a little bit please?

Robert.

Both theoretical and empirical considerations are associated with higher-order CFA. Higher order CFA can be thought of a explicitly representing the causal constructs that impact the first order factors. In other words, the first order factors acts as indicators of the second order factors.

Use of first order or second order scale depends on the research objective.   For example:  we  may examine patient satisfaction  either as  a first order or second order construct (model).  

A first-order construct has observed variables (i.e., the items in its measure) as indicators of the construct.  In first order model we consider  patient satisfaction as a latent construct ( unobserved) and measure it using direct measurable items like patients’ expectations are met, satisfaction with regard to clinical quality, satisfaction with physicians’ behaviour, satisfaction with nursing staff etc.  In second order model , we further consider latent dimensions of patient satisfaction. For example: if we consider patient satisfaction as a function of physicians behaviour, nursing behaviour, administrative functioning,  tangibility etc. then each of these dimensions need to be measured using direct measurable items.

Your second question is not clear. Construct validity that includes convergent validity and discriminant validity need to be established if our interest is in scale development.

Best wishes.

Hi, maybe you need a theoretical support but sometimes there is not any previous literature and it make sense to use a second-order construct as an alternative model (for example the theoretical model are the corelated first-order constructs where there is some theory supporting the relatioship; and the alterative model will be a second-order construct that reflects on the first-order dimensions). Convergent validy problems are not usually solved by introducing a second-order construct since it depends maybe on the items that compose/reflect on the construct. However you can solve the lack of discriminant validity.