I am constructing a scale. I have 80 items and 300 sample data. I ran EFA - Principal Axis Method of extraction and direct oblimin rotation in SPSS. I have 23 factors (using both Eigen value and parallel analysis method). After ignoring items loaded less than 0.3, I have 18 factors. I would like to reduce the constructs further down. 1. Will second order factor analysis will help? I took the total score of each factor for the same 300 sample like F1, F2 ... F18 and ran another EFA in SPSS. I now have 4 factors. Is this method appropriate and in practice? 2. Can I run CFA for these 4 factors for confirming model fit using Chi Square 1 to 3, GFI>0.9, CFI>0.9, and RMSEA<0.08 as criterias?