演講公告
下個星期五有兩場演講
我想 大家可以選擇有興趣的去聽吧！ 畢竟 要寫心得了...
如果我又睡過頭 記得call我阿 謝謝
清華大學、交通大學
統 計 學 研 究 所
專 題 演 講
題 目: Universal Process Capability Index – UPC
主講人: Dr. Shin Ta Liu (Lynx Systems)
時 間: 97年3月7日（星期五）10：00 - 10：50
（上午10：50-11：10茶會於統計所821室舉行）
地 點: 清大綜合三館837室
Abstract
The process capability CPK provides a metric to measure the process leverages
relating to the process specifications. It is an important yard stick in the
six sigma implementations. However, CPK does not necessary related to the
none-conformance rate as six sigma practitioners have claimed. The problem is
that in the CPK formula which projects two parameters space (location and
dispersion variables) into a single parameter space (CPK). As a result of
this shortcoming, two different processes with same CPK can have very
different none-conformance rates, and on the other hand, the processes have
different CPK can have the same none-conformance rate.
In this paper/talk, an alternate process index UPC (Universal Process Index)
is proposed. It is none parametric, so the underline distribution of the
normality is not required. The only assumption for the distribution of the
process parameter is the cumulative distribution function (CDF) of the
process parameter F is monotonically increasing. It is universal, since it
allows a direct comparison of two very different processes.
The estimation of the UPC is in the domain of the statistical estimation of
the tail probability of a CDF F (so is the reliability). None parametric
estimation of the UPC using ordered statistics is given, A Monte Carlo
simulation of the proposed algorithm indicates this method can be reliably
used in the practical situations. The UPC plot also proposed to provide a
visual comparison of multi-processes.
敬請公佈 歡迎參加
清華大學、交通大學
統 計 學 研 究 所
專 題 演 講
題 目: High-Dimension, Low-Sample Size Perspectives in Constrained
Statistical Inference: The SARSCoV RNA Genome in Illustration
主講人: 蔡明田 博士 (中研院統計所)
時 間: 97年3月7日（星期五）11：10 - 12：00
（上午10：50-11：10茶會於統計所821室舉行）
地 點: 清大綜合三館837室
Abstract
High-dimensional categorical data models, often, with inadequately large
sample sizes, crop up in many _elds of application. The SARS epidemic,
originating in Southern China in 2002, had an identi_ed single-stranded and
positive-sense RNA virus with large genome size and moderate mutation rate.
This genomic study is used as a prime illustration for motivating appropriate
statistical methodology for comprehending the genomic variation in such high
dimensional categorical data models. Because of underlying restraints, a
pseudo-marginal approach based on Hamming distance is considered in a
constrained statistical inference setup. The union-intersection principle and
jackkni_ng methods are incorporated in exploring appropriate statistical
procedures.
敬請公佈 歡迎參加
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 59.115.224.61
討論串 (同標題文章)