If your primary objective is to identify the latent dimensions or constructs represented in the original variables, then common factor analysis might be more appropriate than principal component analysis.

In terms of the sample size, it should be fine to run a factor analysis. However you need to assess factorability of the data (The Bartlett test of sphericity and Kaiser-Meyer-Olkin (KMO) Test for Sampling Adequacy), it will tell you whether factor analysis could be applied to your data.

It is advised that exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) shouldn't be performed on the same sample. Alternatively, also it's not encouraged you can randomly split your sample in two, and perform EFA on one split, and CFA on another.

Worthington, R.L. and Whittaker, T.A., 2006. Scale development research: A content analysis and recommendations for best practices. The Counseling Psychologist, 34(6), pp.806-838.

Dear Athina,

it seems that your goal is to develop a broad composite of specific resilience factors. Theoretically, such a composite represents an "aggregate construct" (I like that better than the ambiguous "formative construct" term). Hence, such a scale does not have validity, because you don't measure anything (see the work by Borsboom et al. on the difference between aggregates and latent variables for validity concerns)

I concur with Athina Daniilidou to first focus on the essential ingredients of the later composite, as these ingredients can be perceived as latent variables and their indicators as measures of these latent variables. Hence, these measures can be judged with a validity-lense. And yes, a CFA may be the way to go (as you seem to have some theoretical background in resilience research, I guess you can come up with a theoretical proposal what "resilience" should encompass. Once you have identified 2-3 valid, strongly loading indicators for each resilience latent, you can move forward to select the 2 or even 1 best/strongest indicator for your composite.

That having said, the question remains what the problems of using composites beyond a the focus on a bunch of specific isolated (but perhaps correlating) resilience latents are...

With regard to your question, your N is not the problem. Measurement and scale development are a matter of a) ontology (what is the entity?) and b) causality (what is the relationship between the measure and the construct?)--not statistics.

Best,

Holger

Borsboom, D., Mellenbergh, G. J., & van Heerden, J. (2004). The concept of validity. Psychological Review, 111(4), 1061-1071.

Borsboom, D., Mellenbergh, G. J., & van Heerden, J. (2003). The theoretical status of latent variables. Psychological Review, 110(2), 203-219.

That sounds fine....as a rule of thumb 3-5 x the number of items gives you a workable sample size for FA.

Nigel Marlow Holger Steinmetz Artyom Golossenko Thank you so much for your time and your answers. I will examine your recommendations thoroughly.