Good day,

As the question above implied, can we delete for than 20% items of the whole model because we still have more poor factor loadings? This would mean that the theoretical foundation and the model are not consistent, because the model shows quite a number of poor factor loadings, and if I deleted maximum 20% of the model, I have have the lowest factor loadings around .20++. I wonder if it is acceptable to delete more item, because what Hair et al. (2010) suggested are quite conflicting, such as deleting 20% of the model and recommending to test on new sample if needed to delete more, but then said that the minimum item required in a construct is 3. Which one should I follow?