IteratedBucklin¶
-
class
svvamp.IteratedBucklin(population, **kwargs)[source]¶ Iterated Bucklin method.
Inherits functions and optional parameters from superclasses
ElectionResultandElection.Example: >>> import svvamp >>> pop = svvamp.PopulationSpheroid(V=100, C=5) >>> election = svvamp.IteratedBucklin(pop)
The candidate with least adjusted median Borda score (cf. below) is eliminated. Then the new Borda scores are computed. Etc. Ties are broken in favor of lower-index candidates: in case of a tie, the candidate with highest index is eliminated.
Adjusted median Borda score:
Let
med_cbe the median Borda score for candidatec. Letx_cthe number of voters who put a lower Borda score toc. Thenc‘s adjusted median ismed_c - x_c / (V + 1).If
med_c > med_d, then it is also true for the adjusted median. Ifmed_c = med_d, thenchas a better adjusted median iffx_c < x_d, i.e. if more voters give tocthe Borda scoremed_cor higher.So, the best candidate by adjusted median is the
Bucklinwinner. Here, at each round, we eliminate the candidate with lowest adjusted median Borda score, which justifies the name of “Iterated Bucklin method”.Unlike Baldwin method (= Iterated Borda), Iterated Bucklin does not meet the Condorcet criterion. Indeed, a Condorcet winner may have the (strictly) worst median ranking.
CM(): Non-polynomial or non-exact algorithms from superclassElection.ICM(): The algorithm from superclassElectionis polynomial and has a window of error of 1 manipulator.IM(): Non-polynomial or non-exact algorithms from superclassElection.not_IIA(): Non-polynomial or non-exact algorithms from superclassElection.TM(): Exact in polynomial time.UM(): Non-polynomial or non-exact algorithms from superclassElection.-
candidates_by_scores_best_to_worst¶ 1d array of integers. Candidates are sorted according to their order of elimination.
By definition / convention,
candidates_by_scores_best_to_worst[0]=w.
-
scores¶ 2d array of integers.
scores[r, c]is the adjusted median Borda score of candidatecat elimination roundr.
-
w¶ Integer (winning candidate).
-