To study neutrino oscillations, the knowledge of the initial neutrino energy is required. This energy cannot be determined directly because neutrino beams have a broad energy distribution. Instead, the initial energy for each event is estimated from the final state particles of a neutrino-nucleus interaction using two main approaches. It can be determined either from the total energy of all the final state particles or, if the neutrino scatters quasi-elastically from a bound nucleon, then the initial energy can be calculated approximately using the scattered angle and energy of the outgoing charged lepton. This requires a detailed understanding of neutrino-nucleus interaction cross section for various interaction channels, for different atomic nuclei, and for a wide range of neutrino energies. None of these energy reconstruction techniques have been tested experimentally using beams of known energy. We exploited the similarity of electron-nucleus and neutrino-nucleus interactions, and applied the methods of neutrino energy estimation to Jefferson Lab CLAS electron scattering data for 1.1, 2.2 and 4.4 GeV electrons incident on 3He, 4He, C and Fe targets. We show that the energy reconstruction from the scattered electron plus proton provides a better description of the beam energy than the energy reconstruction from the scattered electron alone, however only 16 - 55% of events reconstruct to within 5% of the beam energy. The energy reconstruction works better for lighter nuclei and lower energies. The tails in the reconstructed energy distributions, corresponding to low (compared to the beam energy) reconstructed energy values, are almost identical for the two reconstruction methods and the consistency between the results from two reconstruction methods does not imply accuracy. We show that energy reconstruction is improved by restricting the transverse momentum of the scattered electron plus proton to be smaller than the nuclear Fermi momentum. We have also compared the results from data to GENIE neutrino event generator results running in electron scattering mode, using the most up to date version. We show that GENIE fails to fully describe the data. We also looked at the potential effects of incorrect energy reconstruction for the proposed DUNE experiment, by generating events with GENIE but reconstructing them using our data. The difference is far greater than the needed DUNE precision.