The $Z$-transformation fitting technique purports to be the best choice for
this type of extraction due to the true functional form of $G_E^p$ being
mathematically guaranteed to exist within the parameter-space of the fit
function. In this article, we examine the mathematical bias and variances
introduced by choosing this technique as compared to the more traditional $Q^2$
fits to directly test if it is truly a better technique.
A selection of $G_E^p$ parameterizations with known fit radius were
statistically sampled and fit with in both $Q^2$ and in $Z$. The mean and
variance of the extracted radii were compared to the input radius. The fits
were performed both with the parameters unbound and bound to study the effect
of choosing physically-motivated bounds. Additionally, the data are fit with
Rational(N,M) functions in $Q^2$ as used by the PRad collaboration.
As expected, the results in both $Q^2$ and $Z$ were poor for unbounded
polynomials, with the fits in $Q^2$ showing a rather large bias with small
variance while the fits in $Z$ had a small bias with a large variance. The
application of bounds yield small improvements. In most cases, fits in $Q^2$
with Rational(N,M) type functions yielded improved results.
We find that the $Z$-transformation technique is a useful tool, but not a
universal improvement over fitting in $Q^2$. The "best" technique was dependent
on the parameterization being fit and the $Q^2$ range of the data. The fits
with a Rational(N,M) function in $Q^2$ typically provided improved results over
other techniques when the data has sufficient $Q^2$ range, but provided
unphysical results when the range was too small. We conclude that any planned
fit should make use of psuedo data to determine the best basis and functional
form for a given data set. In the case of new experiments, this selection
should ideally be done prior to the collection of the data.