|
Abstract:
|
To estimate the number of unique valid signatures on a petition, the authors have recently shown that a new non-linear estimator, W, is preferably to the more commonly used Goodman-type estimator provided that a signature's being replicated on the petition is independent of the signature's being invalid for other reasons, such as not a registered voter or wrong address. Through an extensive enumeration, this paper investigates the power of the Chi squared test of independence to determine empirically the preferred estimator for a petition with known levels of petition size, sample size, percent replicated signatures, percent invalid signatures, and degree of association between replicated and otherwise invalid signatures.
|
