Big Data Wharton Wav 9 2012 V3
Download as PPTX, PDF
0 likes303 views
This document summarizes a presentation about analyzing large datasets and drawing conclusions from evidence. It cautions that the analyst's assumptions can influence their analysis and that solid methodology is important. Relationships may be missed if analysis focuses only on superficial measures. Presenting findings clearly is also important, as counterintuitive conclusions require strong evidence and elegant insights can be undermined through poor visualization.
Big Data Wharton Wav 9 2012 V3
- 1. Big Data, A Material Issue Wharton Aerospace San Francisco CA 27 September, 2012 W. Alexander Vacca, Ph.D. Corporate Director- Business Assessment Northrop Grumman Corporation The views expressed in this presentation are solely those of the presenter, and do not necessarily reflect those of Northrop Grumman, its employees, its customers, or its shareholders. Please do not cite without permission
- 2. Disclaimer The views expressed in this presentation are solely those of the presenter, and do not necessarily reflect those of Northrop Grumman, its employees, its customers, or its shareholders. Please do not cite without permission 2
- 3. Big Data: A Material Issue -or- What a great 1990s post punk proto alternative power pop trio from Chicago can teach us about drawing conclusions from evidence 3
- 4. Assumptions of the Analyst Shape Conclusions Well I know it's been a little while since the first time that I made you smile and I thought I'd take this opportunity to dig a little deeper inside and try to find out what someone like you is trying to hide from me. Well I hope that you don't get angry girl, and I hope that you don't get hurt by the things I found while goin' through your purse. 4 Material Issue Goin Through Your Purse, Freak City Soundtrack, (Polygram Records, 1994) Lyrics by Jim Ellison
- 5. Assumptions of the Analyst Shape Conclusions The assumptions of the analyst condition the choice of data sets, the choice of metrics, the hypothesized relationships to be tested, and the interpretation of the evidence. Artillery casualties in the Russo-Japanese War (1904-05) Intense and good faith effort to quantitatively measure the effects of heavy artillery on modern warfare by tabulating data from field hospitals Relatively few fatal and non-fatal wounds recorded from artillery fire Therefore artillery did not dramatically change battle tactics BUT most artillery casualties never made it to field hospitals (missing in action) AND artillery pinned troops, allowing opposing infantry to close and inflict wounds with rifles Artillery did change battle tactics, as the observing militaries discovered in 1914 5
- 6. Russo Japanese War Casualty Analysis Per cent. Japanese casualties from infantry fire .. 91.35 Japanese casualties from artillery fire .. 7.99 Japanese casualties from bayonet, saber, etc ... .66 ---------- Total 100.00 John Morrison report, page 83, United States War Department: Office of the Chief of Staff (Military Information Division) Reports of Military Observers Attached to the Armies in Manchuria During the Russo Japanese War (volume 1) (US Government Printing Office, September 1906) 6
- 7. Making Sense of Data Your make-up, and your compact, and your lipstick, and stuff like that, and the keys that you keep on a chain. Your cigarettes and my lighter too, and a picture of your sister, God I think youre one in the same. Well a photo of your mother too, and all the boys who dated you, and your high school graduation ring. And a check stub from the place you work, and some poetry from some stupid jerk who's trying to steal your heart from me. 7 Material Issue Goin Through Your Purse, Freak City Soundtrack, (Polygram Records, 1994) Lyrics by Jim Ellison
- 8. Making Sense of Data Big data is not large-n, It is a relational constellation of rich and diverse elements with different pedigrees from which probabilistic predictive conclusions can be drawn. Solid methodology, including thoughts about hypotheses, need to underpin our analysis. There is no substitute for critical thinking and statistical literacy. Slick tools poorly applied can still deliver convincing results. We increase our level of confidence in junk conclusions by paying insufficient attention to methods and assumptions. Beware easy fast tools. You can cut off your fingers faster with an electric band-saw than with a manual hack saw. 8
- 9. Presentation of Findings Matters Then I cleverly put all these things back into place so you wouldn't think that I was a jerk. Yeah the things I found while goin' through your purse. Well I hope that you don't get offended girl, yeah and I hope you don't get hurt, by the things I found while goin' through your purse. 9 Material Issue Goin Through Your Purse, Freak City Soundtrack, (Polygram Records, 1994) Lyrics by Jim Ellison
- 10. Presentation of Findings Matters Counterintuitive conclusions face a higher burden of proof, and some of the most valuable insight is counterintuitive. Relationships may be missed because they are masked by superfluous graphical or presentational artifacts Elegant conclusions can be undermined by poor graphics This Powerpoint deck is an example of this problem- a complex and multifaceted argument has been reduced to a series of linear bullet points (Arent you glad youre here for the briefing?) Everybody should read or listen to Dr. Edward Tufte (www.edwardtufte.com) 10