Hello Ahmed,

Given that the factor structure of the SGRQ appears to be unstable (i.e., it differs from sample to sample), I think that you need to conduct only EFA. CFA is not available through SPSS, anyway.

Other people might disagree with me, so I hope you are given advice from people other than me.

All the best with your research.

As you said, most validation studies on SGRQ have focused on convergent and discriminant validity and only a limited number of studies have explored factor analysis, your plan to conduct both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) appears well-grounded.

Necessity of EFA and CFA: EFA is useful for exploring the potential underlying factor structure without imposing a priori constraints. This can be particularly useful if previous studies have not definitively established the number of factors or have found varying structures.

CFA (use AMOS) allows you to test how well the data fits a predefined factor structure based on previous research or theories. Since some studies have already attempted this with varying results, conducting a CFA can provide a comparative analysis against these established structures.

Conducting both EFA and CFA will enable you to first explore the data freely (EFA) and then test a specific model structure (CFA), providing a robust analysis of the factor structure of the SGRQ in your sample.

Pallant, J. (2013). SPSS Survival Manual. McGraw-Hill Education. - This book offers a practical guide to using SPSS for EFA and CFA, suitable for handling SGRQ data.

Kline, R. B. (2015). Principles and Practice of Structural Equation Modeling. Guilford publications. - This text provides a comprehensive overview of structural equation modeling, including both EFA and CFA, and discusses when and why each method should be used.

Ahmed Shiyam Absolutely, conducting Exploratory Factor Analysis (EFA) before Confirmatory Factor Analysis (CFA) can indeed strengthen your model validation process. Regarding software, yes, SPSS Amos can be a suitable tool for conducting both EFA and CFA.

Thank you so much for your quick reply. Sasidharan Sivakumar Dr Amit Kumar Robert Trevethan

Ahmed Shiyam,

Hello again, Ahmed. I realise that I could have added something to my answer up above. Under the circumstances, I still think you would need to conduct only EFA (not CFA), but if - for some reason - you wanted to conduct CFA, you should do that first using the combination of items and factors that you regard as being most reasonable. If the CFA was unsuccessful, you could then go on to conduct EFA to identify what "went wrong". For the initial CFA and subsequent EFA, you could use the same data.

You should not work the other way around, however, i.e., conduct EFA first and then, on the basis of your results and with the same data, conduct CFA. Doing things that way around is essentially a self-fulfilling prophesy (and useless).

Again, all the best with your research.

To everyone who has contributed to this thread so far:

I confess that, when I first saw Ahmed's question, I assumed that the instrument he was referring to had a Likert-type format with at least five response options for each of the 50 items.

I have since discovered that the 50 items have a variety of formats, some of which have dichotomous responses (true/false), some of which have three response options, some of which have four, and some of which have five.

I believe that such a variety of response options would invalidate any attempts at EFA and CFA. Do others agree with me concerning that? I hope that, together, we can offer Ahmed good advice.