Dear Markus Schulze, thanks for your reply. Basically, I have to come up with some method or way to select one of the two rankings you gave for A10, A12, A23, A33, A67. That is a real problem.
Maybe we could use approval voting in this case to reduce the number of candidates (hopefuls), I don't know. These things happen in regression too when there are too many candidate variables and too few data (forgot what it is called), therefore it is standard to add a requirement that each variable (candidate) should be significant in a univariate model at, say 5% or so in order to qualify as a candidate in the multivariate model. This requirement could for instance translate to a requirement of eliminating the first X candidates, with the lowest Schulze-single winner ranking (or the lowest share of preferences of the candidate on first M places of the ballot) in the Schulze-STV election. That heuristic could eliminate the candidates before they enter the model and might resolve the ambiguities in some of the elections you describe. R Fobes mentioned using approval voting to pre-select candidates. I would like to send you an input file, but first I have to generate some test-data. I could start with some fictive elections, since we don't use ranked ballots in our party. A full-scale test will take some time to arrange. By the way, out of pure curiosity, could a hybrid ranked ballot, i.e. a ballot on the form A=B>C=D>E, be used in Schulze-STV in theory, without sacrificing any of the good properties of the method? Best regards Peter Zborník On Sun, May 9, 2010 at 5:26 PM, Markus Schulze < [email protected]> wrote: > Dear Peter Zbornik, > > you wrote (9 May 2010): > > > In your paper schulze3.pdf, there are some instances, > > where the Schulze proportional ranking fails to produce > > an unambiguous ordering (see for instance the result > > for data set A10). Why do there ambiguities occur and > > how would you recommend them to be resolved in a > > deterministic manner without resorting to random number > > generation etc? > > In 5 instances (A10, A12, A23, A33, A67), the Schulze > proportional ranking is not unique. This is caused by > the small numbers of voters and the large numbers of > candidates. > > For example, in instance A10 (83 voters, 19 candidates), > there are two possible Schulze proportional rankings: > NAPMQFGRSLIBDJKEHOC and NMPQAFGRSLIBDJKEHOC. > > You wrote (9 May 2010): > > > Does Schulze-STV allow for truncated ballots? I.e. when > > there are 5 candidates, does Schulze-STV allow me to > > only rank two of them on my ballot? > > I recommend "proportional completion". > This is explained in section 5.3 of > http://m-schulze.webhop.net/schulze2.pdf > and in the file calcul01.pdf of > http://m-schulze.webhop.net/schulze3.zip > > You wrote (9 May 2010): > > > I am also curious to know, if you think it would be > > difficult for you to implement a program, which would > > handle the green council elections in an optimal > > proportional manner, i.e. methods, which would only > > impose the required ranking. > > It would be simple to incorporate all the requested > specifications. Send me an input file with explanations. > > Markus Schulze > > > ---- > Election-Methods mailing list - see http://electorama.com/em for list info >
---- Election-Methods mailing list - see http://electorama.com/em for list info
