DYNAMIC INTERACTIVE GRAPHICS PEDRO VALERO MORA UNIVERSITAT DE VALÈNCIA I HAVE WORN MANY HATS
• Degree in Psycology
• Phd on Human Computer Interaction (HCI): Formal Methods for Evaluation of Interfaces
• Teaching Statistics in the Department of Psychology since1990, Full Professor (Catedrático) in 2012
• Research interest in Statistical Graphics, Computational Statistics
• Comercial Packages (1995-) Statview, DataDesk, JMP, Systat, SPSS (of course)
• Non Comercial 1998- Programming in Lisp-Stat, contributor to ViSta (free statistical package)
• Book on Interactive Dynamic Graphics 2006 (with Forrest Young and Michael Friendly). Wiley
• Editor of two special issues of the Journal of Statistical Software (User interfaces for R and the Health of Lisp-Stat)
• Papers, conferences, seminars, etc. DYNAMIC-INTERACTIVE GRAPHICS
• Working on this topic since mid 90s
• Other similar topics are visualization, computer graphics, etc.
• They are converging in the last few years
• I see three periods in DIG
• Special hardware
• Desktop computers
• The internet...and beyond
• This presentation will make a review of the History of DIG and will draw lessons from its evolution 1. THE SPECIAL HARDWARE PERIOD
• Tukey's Prim 9
• Tukey was not the only one, in 1988 a book summarized the experiences of many researchers on this topic
• Lesson#1: Hire a genius and give him unlimited resources, it always works
• Or not...
• Great ideas but for limited audience, not only because the cost but also the complexity… 2. DESKTOP COMPUTERS
• About 1990, the audience augments...macintosh users
• Many things only possible previously for people with deep pockets were now possible for…the price of a Mac
• Not cheap but affordable
• Several packages with a graphical user interface:
• Commercial: Statview, DataDesk, SuperANOVA (yes super), JMP, Systat…
• Non Commercial: Lisp-Stat, ViSta, XGobi, MANET... 2.1. COMMERCIAL: DATADESK
• Excellent for Exploratory Data Analysis, founded by P. F. Velleman (student of Tukey) in 1985 (still kicking)
• Linking, selecting, filtering, corkboards, models, fast,
• Teaching
• An interesting niche was teaching, companion of great textbooks on introductory statistics
• However, many people use the textbook with SPSS or other statistical packages.
• Lesson #2: Statistical analysis must be taught with the same package that will be used for real analysis
• It did not succeed although it is still in the market, but not that many people use or speaks about it.
• A mystery to me…reasons?
• Audience? Lesson #3: Think in your audience
• Presentation graphics? Static graphics? Lesson #4: Once the fun is over you need to freeze the picture 2.2.1 NON-COMMERCIAL: VISTA
• Started about 1990 by F. W. Young
• Linking, selecting filtering, spreadplots (like corkboards or dashboards) , statistics, fast...the same as DataDesk but less polished
• Excellent for Exploratory Data Analysis
• Based on XLisp-Stat 2.2.2. NON-COMMERCIAL: LISPSTAT
• Programming language for statistics based on Lisp with elements of S.
• Developed by Luke Tierney, he works currently in R
• Designed specifically for Dynamic Interactive Graphics
• Lisp-Stat was like R/S but with interactive graphics and user interface elements (buttons, lists, mouse events, windows...all native not TCL/TK or Java or Python or else)
• Lisp is great!… according to Ross Ihaka the creator of R
• It was abandoned almost by everybody about 2003
• What did it happen?: In a single letter -> R
• Why?
• Static graphics #Lesson 5: Static graphics (did not I say this before?)
• Familiarity? #Lesson 6: Make great software part of student’s education
• Community? #Lesson 7: You need a community of users and developers 3. THE INTERNET
• The impact of the internet has occurred at many levels
• Colaboration, distribution, spread, has made non-commercial collaborative products very common
• Sometimes is a problem more than a solution: many programs producing solutions for the same problem, projects discontinued, partially developed, etc.
• Delivering products in the browser so everybody can play with them easily
• I do not know much about the commercial products currently in the market (but I would love to know more…any free licenses there?)
• On the non-commercial (libre?), R is the king 3.1 R
• It is great! However, there is an area that could be improved
• Graphical User Interfaces! (Come on, it is 2016!)
• I am talking of something that my colleagues and students in Psychology could use (just copy Statview!)
• No, R-Commander is not sufficiently easy
• TBH, there are some exciting packages that go in the right direction (Shiny):
• However: A good GUI is not only that!
• Lesson #8: Hire specialists in human factors if you wish a really good GUI
• I am pessimistic about getting this right non-commercially, there are things that do not pay off academically DISCUSSION
• Evolution: DIG progressively closer to the masses
• Supporting the idea of data as a public good that anybody can use for understanding the world we live, take decisions, evaluate policies, etc.
• Current barriers steeper on the human factor than on the technical factor
• Testing: What do the people do? What do they understand?
• Education: What do the people know? How can people be taught for understanding statistics?
• Communication: Can the message still be manipulated?
• Cognition: Are human beings able to think statistically? THANKS FOR YOUR ATTENTION