I executed Exlporatory Factor Analysis (EFA) and some of the factors consists of just 2 items. Is it acceptable? I read an article it says that a factor should contain at least 3 items. Is it an obligation?
Could you offer me an article that say a factor can consist of 2 items??? I will need a citation. To explain this situation.
It is recommended in research and statistics books to have at least 3 items per factor or subscale.
As we are exploring the factors in EFA, it is ok to have a factor having two items. We dont have any control on number of items in EFA. But in Confirmatory Factor Analysis (CFA), it is advisable to have at least 3 items per factor. As CFA tests a theoretical model, we should make sure to have at least three items per factor. or it will have an identification problem.
In most cases when using AMOS to run EFA it is advisable to have at least 3 items as an indicators to explain a latent variable, except there is strong theoretical backing for the two items used.
factor
Too add just a small bit to this, if the resulting scale has 2 items it will likely have low reliability (unless the correlation between the two is large) and thus have a problem with replication of the factor across samples.
According to J.H.Hair, each latent construct should contain at least three items to be better explained.
For CFA, If you take two items, the model would be underidentified with negative degree of freedom. In that case, a unique solution cannot be found.
If you take three items per factor, the model would be justidentified with degree of freedom=0. In this case the model would have perfect fit (e.g. CFI/GFI =1). The model is also called saturated chi-square goodness of fit test is zero. Such models are of no or little interest to researchers testing a theory.
A better approach as per my opinion, is to take four items per factor so that a solution can be found with positive degree of freedom and corresponding chi-square, goodness of fit value. It is called as overidentified model.
Reference: Hair, J., Black, W. C., Babin, B. J., & Anderson, R. E. (2010). Multivariate data analysis (7th ed.). Upper saddle River, New Jersey: Pearson Education International.
i have 3,3 items in last 2 factors when i run EFA even CFA. i need reference to retain both factors with 3,3 items.can you recommend any reading material?
Gökhan Aksu, Mehmet Taha eser, Cem Oktay Güzeller: Açımlayıcı ve Doğrulayıcı Faktör Analizi ile Yapısal Eşitlik Modellemesi (s. 5)
Hocam bu kitabı temin edin ve sayfa 5'i inceleyin. Faktör 1 ya da 2 maddeden oluşuyorsa korelasyon katsayısı .70 üzerinde ise kabul edilebilir. Sevgiler
- Badges
- Science topic
- Similar topics
- Psychology