교락(confounding) -자료 부족으로 인해 인자의 개별 효과를 알 수 없는 경우


역인과관계의 오류확인법

"경찰관의 수가 많을수록 범죄 발생 건수가 늘어난다"

는 말을 시험하기 위해 경찰관의 수를 줄여 범죄 발생건수가 주는지 확인 할 수 있다.


교호작용(interaction)

어떤 결과변수에 영향을 미치는 독립변수가 2개 이상일때 이 독립변수가 서로 교호작용을 할 수 있다.

즉, 모뎀과 컴퓨터 기종에 따른 처리속도에 예에서, 모뎀 용량에 따른 처리속도가 컴퓨터 기종에 영향을 받지 않으면 교호작용이 없다고 말하며, 처리속도가 컴퓨터 기종에 영향을 받으면 교호작용이 있다고 말한다

즉, 한 인자의 효과가 다른 인자의 조건에 따라 달라지는 것이다


indicator function(지시함수)

특정집합에 특정 값이 속하는지를 표시하는 함수로 특정 값이 집합에 속한다면 1 속하지 않는다면 0으 ㅣ값을 가짐

characteristic function도 같은 뜻인듯..


분산팽창계수(VIF:Variance Inflation Factor)

회귀 분석시 예측 변수 간의 상관관계를 알아보기 위하여 사용, 만약 X들간의 상관관계가 크다면, 결과가 안정적이지 않음

VIF가 10보다 크면 예측 변수중 적어도 2개는 상관관계가 매우 크므로, VIF가 안정될 때까지 VIF가 가장 큰 변수부터 하나씩 순차적으로 제거해보면서 활용해야 한다.

R에서는 vif함수로 사용가능


'Mathmatics > Statistics' 카테고리의 다른 글

기하분포, 이항분포, 푸아송 분포 완전 정리  (0) 2014.03.16
수학기호 "~"  (0) 2013.04.09

500 TB Or More Of Data Under Management, According To New InformationWeek Reports Research

Yet just 22% say data analysis is core to how they do business.

Published: Thursday, Nov. 29, 2012 - 9:09 am

/PRNewswire/ -- UBM Tech's InformationWeek Reports, a service provider for peer-based IT research and analysis, announced the release of its latest research report. 6 Big Data Lies encompasses analysis of results from InformationWeek's recent Big Data Survey shares with readers best practices and pitfalls when collecting and analyzing data.  Two-hundred and fifty-seven respondents participated in the survey.

(Photo: http://photos.prnewswire.com/prnh/20121129/SF20332-INFO)

Research Summary:

The ability to collect and analyze data is a core IT value proposition and there's no question there's more data available than ever, especially information from the Web and the new wave of smart mobile devices.  The issue, though, is that most businesses aren't good at using the data they have now.  What are the odds companies will get better with more being piled on? More data doesn't fix bad analysis.

Findings: 

  • 45% of respondents plan to grow analytics tools capabilities over the next 12 months.
  • 35% include CRM in their integrated data analysis and reporting programs, but only 29% include email and just 22% include Web analytics.
  • Only 20% plan to increase their dedicated analytics staff over the next year.
  • Just 9% say their organizations are very effective at identifying critical data and using it to make decisions.

The report author, Michael Healey, is the president of Yeoman Technology Group, an engineering and research firm focusing on maximizing technology investments for organizations, and an award-winning InformationWeek Reports contributor.

For full access to the research data, members can download now:http://reports.informationweek.com/abstract/81/9081/Business-Intelligence-and-Information-Management/research-big-data-smart-data.html?cid=rpt_press_rls

"Companies have been doing an adequate job in building the infrastructure needed for data analysis," says Lorna Garey, content director of InformationWeek Reports. "However, our research shows there are major gaps in data analysis, even for the most common pieces of information that have been in use for years, such as transactional data, system logs, email, CRM, even Web analytics."

For more information: Art Wittmann     VP & Managing Director, InformationWeek Reports 415-947-6361 art.wittmann@ubm.com

About UBM Tech

UBM Tech is a global media business that provides information, events, training, data services, and marketing solutions for the technology industry.  Its media brands and information services inform and inspire decision makers across the entire technology market— engineers and design professionals, software and game developers, solutions providers and integrators, networking and communications executives, and business technology professionals.  UBM Tech's industry-leading media brands include EE TimesInteropBlack HatInformationWeekGame Developer ConferenceCRN, and DesignCon. The company's information products include research, education, training, and data services that accelerate decision making for technology buyers.  UBM Tech also offers a full range of marketing services based on its content and technology market expertise, including custom events, content marketing solutions, community development and demand generation programs. UBM Tech is a part of UBM (UBM.L), a global provider of media and information services with a market capitalization of more than $2.5 billion.

SOURCE UBM Tech



출처 :http://www.sacbee.com/2012/11/29/5018763/45-of-companies-have-500-tb-or.html


Read more here: http://www.sacbee.com/2012/11/29/5018763/45-of-companies-have-500-tb-or.html#storylink=cpy

+ Recent posts