Precision experiments, such as those performed at LEP and SLC, offer us an excellent opportunity to constrain extended gauge model parameters. To this end, it is often assumed, that in order to obtain more reliable estimates, one should include the sizable one--loop Standard Model (SM) corrections, which modify the $Z^0$ couplings as well as other observables. This conviction is based on the belief that the higher order contributions from the ``extension sector'' will be numerically small. However, the structure of higher order corrections can be quite different when comparing the SM with its extension, thus one should avoid assumptions which do not care about such facts. This is the case for all models with $��_{\rm tree} \equiv M_W^2/(M_{Z}^2\cos^2{��_W}) \neq 1$. As an example, both the manifest left-right symmetric model and the $SU(2)_L \otimes U(1)_Y \otimes \tilde{U}(1)$ model, with an additional $Z'$ boson, are discussed and special attention to the top contribution to $����$ is given. We conclude that the only sensible way to confront a model with the experimental data is to renormalize it self-consistently, if not, parameters which depend strongly on quantum effects should be left free in fits, though essential physics is lost in this way. We should note that arguments given here allow us to state that at the level of loop corrections (indirect effects) there is nothing like a ``model independent global analysis'' of the data.

10 pages