I have a question about the CASSCF method. According to what I have read, it is easiest to build the CASSCF active space by selecting specific localized orbitals from a small-basis DFT or MP2 calculation. With near-minimal basis sets, it is easy to identify important active space orbitals visually (e.g., $\sigma$, $\sigma^*$, etc.)
For large basis sets, it appears to be straightforward to visually identify the active space orbitals that originate from occupied localized orbitals. However, identifying active space orbitals that originate from unoccupied localized orbitals (e.g., $\sigma^*$) is more difficult because of the large number of virtual orbitals.
Does this mean CASSCF calculations are not run with orbitals computed from large basis sets?