포스터랑 논문<-- 제 개인 공유 droplr에 링크 걸었습니다. 

"모바일을 생각못했네요;;"

각각 pdf파일입니다.

___________

여러 커뮤니티에서 논란이 되고 있는 K값 관련된 것과, 시나리오 관련한 것에 대해 설명이 충분히 들어 있나 저도 좀 더 읽어봐야 겠네요 ;-)


두가지 시나리오에 대해 이야기를 하는데.

Scenario 1: unintentional systematic bias
Assume a consistent op-scan performance
across the country:
P2 ~ B(P, r+β), where M2 ~ B(M, r).

Scenario 2: intentional systematic manipulation
Apply conditional Bernoulli and multinomial distributions to sorting algorithm with K=1.5.

Scenario 1 resulted in an overall mismatch of the variability of unclassified rates observed at district level, whereas scenario 2 returned comparable results to the actual election outcomes at both district and national levels.

결론은 

This study demonstrates a potential serious loophole in using the op-scan counters, which can be error-free but not manipulation-free
Op-scan counters can generate serious misclassifications by a pre-programmed algorithm, not by random mechanical malfunctions.
Serious errors can go undetected if results are not audited effectively, resulting in election winner change.

대충 내용은 이거인듯 합니다...간만에 통계논문이나 읽어보고..자야겠군요;;

나머지 내용은 읽어보고서 ;-)