0
$\begingroup$

I have a dataset with countries as observations and 33 country-level continuous variables.

The information accompanying the dataset groups the 33 variables into 6 different 'categories', however I am not convinced that this categorization is correct as some variables from different categories are highly correlated with with each other.

I beleve that some variables have been wrongly lumped together within these categories and that by performing CFA I would be able to remap the variables onto new categories.

However, I have only seen EFA and CFA applied to survey data. Does it make sense to perform CFA on country-level continuous variables or would PCA be more appropriate here?

Improve this question
$\endgroup$

2 Answers 2

Reset to default
1
$\begingroup$

CFA can absolutely be used on country-level data. One of the most famous examples of CFA, Bollen's political democracy analysis, is on country-level data. See Bollen (1980).

Improve this answer
$\endgroup$
1
  • $\begingroup$ Thank you for your answer and for pointing me towards the article! $\endgroup$
    – Hamson
    Jan 6, 2020 at 14:39
1
$\begingroup$

Yes, you can do CFA with cases that are countries (or companies, or...). Be aware that the variables may have been grouped together for conceptual reasons, rather than because of their statistical behavior, so rejecting a factor model will not invalidate the grouping. Moreover, n = 33 is a very small sample size for CFA. With 33 observed variables, your model will likely have more than 70 parameters to be estimated, which will make any results obtained highly suspect.

Improve this answer
$\endgroup$
3
  • 1
    $\begingroup$ I also initially thought they had a sample size of 33, but I don't think that is implied by their question. $\endgroup$
    – Noah
    Jan 6, 2020 at 8:21
  • $\begingroup$ Yes, my mistake. Sigh. $\endgroup$
    – Ed Rigdon
    Jan 6, 2020 at 11:28
  • $\begingroup$ Thanks for the answer. The number of variables is 33 and my sample size is n = 4416 (100+ countries with over 20+ years each). The variables have been grouped together for conceptual reasons, however I believe that some of the groupings are not as 'sound' as others. $\endgroup$
    – Hamson
    Jan 6, 2020 at 14:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.