The response of the proton to elastic scattering events has long been known to be described via two functions of the squared momentum transfer: the Sachs electric and magnetic form factors Ge and Gm, respectively. To understand the elastic structure of the proton there are two main observables which constrain the form factors: unpolarized cross section data from elastic electron-proton scattering events, and polarization transfer measurements, which generate polarization ratio data. After taking into account tree-level radiative corrections, separate fits of the form factors to these data lead to fits which disagree significantly with one another. Higher loop corrections, especially two-photon-exchange corrections, have been assumed to play a larger role than previously thought in order to explain this discrepancy. We present our global reanalysis which takes special care to treat normalization uncertainties in a most statistically rigorous manner, with additional work done to understand how to extend certain statistical methods only defined for linear models, to non-linear models. We find only minor differences to fits when normalization uncertainty is correctly accounted for. We also present a simultaneous global fit to the two-photon-exchange corrected cross section data and polarization ratio data. In addition to the above, an independent recalculation of the resonance contribution to the F3gz structure function is computed using quark model helicity amplitudes.