Presentation is loading. Please wait.

Presentation is loading. Please wait.

Data visualization and graphic design Introducing R for data visualization Allan Just and Andrew Rundle EPIC Short Course June 21, 2011 Wickham 2008.

Similar presentations


Presentation on theme: "Data visualization and graphic design Introducing R for data visualization Allan Just and Andrew Rundle EPIC Short Course June 21, 2011 Wickham 2008."— Presentation transcript:

1 Data visualization and graphic design Introducing R for data visualization Allan Just and Andrew Rundle EPIC Short Course June 21, 2011 Wickham 2008

2 Intro to R Objectives After this class, participants will be able to: 1.Describe some capabilities and uses of R 2.Search for help within R and use good coding practices for reproducible research in R 3.Read in and summarize a simple dataset with R/JGR/Deducer 4.Make some standard plots with Deducer templates

3 What is R? nytimes.com

4 R has many uses Work with data: subset, merge, and transform datasets with a powerful syntax Analysis: use existing statistical functions like regression or write your own Graphics: graphs can be made quickly during analysis and polished for publication quality displays

5 Why learn a whole language to look at data versus Excel? 1.Recreate/redo your exact analysis 2.Automate repetitive tasks 3.Access to statistical methods not available in Excel 4.Graphs are more elegant

6 1.It's free! 2.It runs on Mac, Windows, and Linux 3.It has state-of-the-art graphics capabilities 4.It contains advanced statistical routines not yet available in other packages – a de facto standard in statistics 5.Can program new statistical methods or automate data manipulation/analysis adapted from statmethods.net Why R versus SAS, SPSS, or Stata?

7 Made in SASRedone in R learnr.wordpress.com

8 R plots from my own research

9 Scatterplot matrix bivariate densities and correlations

10 Forest plot to compare parameter estimates from many models

11 Displaying lots of data: facetted histograms

12 Plotting data with a model

13 Automated report generation

14 Shapefile: CIESIN, Columbia University Asthma data: http://nyc.gov/html/doh/downloads/pdf/asthma/asthma-hospital.pdf Choropleth map

15 Intro to R: recap Objectives After this class, participants will be able to: 1.Describe some capabilities and uses of R Statistical data analysis Automation (scripting) of functions to work with data Elegant graphics to facilitate data visualization 2.Search for help within R and use good coding practices for reproducible research in R 3.Read in and summarize a simple dataset with R/JGR/Deducer 4.Make some standard plots with Deducer templates

16 Learning a new language is difficult flickr.com/photos/dnorman/3732851541/

17 What makes R difficult to learn R is designed to be flexible and powerful rather than simple but limited. R is a fully featured language mainly used from the command line. Learning the commands and the structure of the code takes time and practice. If I made a a typo you would know what I meant...

18 What makes R difficult to learn R is designed to be flexible and powerful rather than simple but limited. The solution: be careful build code in simple pieces and test as you go (learn to debug). Reuse code that works. Use helpful resources. Consider an alternative GUI for R.

19 Getting help in R You can call for help on a function with a leading question mark and leaving off the () ?functionname Search online statmethods.net An Introduction to R in Windows found under Help – Manuals (in PDF)

20 Suggestions for an R workflow Save the bits of your code that work in a text editor - building a script of clean code that works from start-to-finish. With clean code instead of transformed data files it is easier to redo analyses if your data are updated or you want to change an earlier step Leave yourself informative comments # everything to the right of the pound sign # is unevaluated Using spaces and indents can help readability Use meaningful names for objects Reproducible research!

21 Intro to R: recap Objectives After this class, participants will be able to: 1.Describe some capabilities and uses of R 2.Search for help within R and use good coding practices for reproducible research in R ?t.test will bring up R help Free manuals online: Introduction to R Also: statmethods.net #use comments; save the code that works to reproduce your results 3.Read in and summarize a simple dataset with R/JGR/Deducer 4.Make some standard plots with Deducer templates

22 Learning the language Many important features Arithmetic and logical operators: +, <, … Data types: numeric, logical, … Data structures: vectors, matrices, … Functions – always end with (): median(x)

23 Using R as a calculator Mathematical operators + - / * ^ log() abs()

24 R can evaluate logical expressions == equal != not equal & and | or (vertical pipe) 10 < 20 [1] TRUE pi > 3 & 2^2 == 4 [1] TRUE "This" != "That" [1] TRUE

25 Creating new objects Assignment operator is <- (looks like an arrow) x <- 10 “Set x to take the value 10” The symbols in this operator must be adjacent. x < - 10 What does this do? You can overwrite old values x <- x^2 “Set x to take the value x 2 ”

26 R operations are vectorized A vector is an ordered set of data of the same type (numeric, logical, dates, etc.) Concatenate function is c() x <- c(1, 2, 3) x^2 [1] 1 4 9 For integer sequences we can make vectors quickly with n:m 4:9 [1] 4 5 6 7 8 9 Vector recycling c(0,10) * c(1, 1, 1, 1) [1] 0 10 0 10

27 Other object classes Matrices: lots of matrix operations you would always use the lm() function to fit a linear model but you could do it manually beta <- solve(t(X)%*%X)%*%t(X)%*%y Arrays: n-dimensional Lists: sets of objects (can be different classes)

28 Indexing and subsetting Concatenate function is c() x <- c(10, 20, 30) x [1] 10 20 30 Refer to components of objects by a position index which goes between square braces x[2] return the second position in x [1] 20 x[c(1, 2)] return the first and second position in x [1] 10 20 x[-3] return all except the third position in x [1] 10 20 What would x[c(3, 2)] return?

29 Data frames A data frame is a rectangular collection of data Rows: observations Columns: variables diamonds <- data.frame(carat, cut, price) carat cut price 1 0.23 Ideal 326 2 0.21 Premium 326 3 0.23 Good 327 4 0.29 Premium 334 5 0.31 Good 335 6 0.24 Very Good 336

30 Data frames You can extract the variables as vectors with a $ diamonds$cut You can also index by position (or name) with square braces diamonds[2, 3] returns the single value in row 2, column 3 An empty index is treated like a wildcard and corresponds to all rows or columns depending on position diamonds[, "cut"] (same result as diamonds$cut ) How would you return the first three rows and all columns? row, column

31 R functions Thousands of functions are built-in: median()lm() linear model t.test()chisq.test() or make your own: inch.to.cm <- function(x){x * 2.54} inch.to.cm(74) [1] 187.96

32 Missing values These take a value of NA Can be in a data object of any type (logical, numeric, character) By default operations on NA will return NA NA == NA [1] NA Can check for NA with is.na() y <- c(2, 10, NA, 12) is.na(y) [1] FALSE FALSE TRUE FALSE Can often pass na.rm = T option to remove NA values in operations mean(y) [1] NA mean(y, na.rm = T) [1] 8

33 R has several thousand additional packages time series survival spatial machine learning bioinformatics Interfaces to Excel, SQL databases, Twitter, google maps…

34 Installing a package 1.Open up R 2.Click in to the console window and type: install.packages() 3.Select a mirror (anywhere in the US) 4.Find and select "Deducer" and choose OK. 5.This will download Deducer and the other packages which it requires, including ggplot2.

35 The default R graphical user interface (Windows)

36 JGR

37 Deducer

38 Recap on GUIs R Default Windows GUI: lacks additional features to make learning or programming easier JGR: Makes programming easier with syntax highlighting and command argument suggestions. No menus for stats. Looks the same across platforms (Java based) Deducer: Adds menus for basic stats to JGR. Menu driven graphics options (building with ggplot2).

39 R graphics – 3 main "dialects" Base: with(airquality, plot(Temp, Ozone))Lattice: xyplot(Ozone ~ Temp, airquality) ggplot2: ggplot(airquality, aes(Temp, Ozone)) + geom_point( )

40 Google image search: ggplot2

41 ggplot2 philosophy Written by Hadley Wickham (Rice Univ.) Extends The Grammar of Graphics (Wilkinson, 2005) All graphs can be constructed by combining specifications with data (Wilkinson, 2005). A specification is a structured way to describe how to build the graph from geometric objects (points, lines, etc.) projected on to scales (x, y, color, size, etc.)

42 ggplot2 philosophy When you can describe the content of the graph with the grammar, you don’t need to know the name of a particular type of plot… Dot plot, forest plot, Manhattan plot are just special cases of this formal grammar. …a plotting system with good defaults for a large set of components that can be combined in flexible and creative ways…

43 Building a plot in ggplot2 data to visualize (a data frame) map variables to aesthetic attributes geometric objects – what you see (points, bars, etc) scales map values from data to aesthetic space faceting subsets the data to show multiple plots statistical transformations – summarize data coordinate systems put data on plane of graphic Wickham 2009

44 A basic ggplot2 graph ggplot(airquality) + geom_point(aes(x = Temp, y = Ozone)) Data Aesthetics map variables to scales Geometric objects to display

45 A ggplot2 graph is an R object p <- ggplot(airquality) + geom_point(aes(x = Temp, y = Ozone)) str(p) #structure of p List of 8 $ data :'data.frame':153 obs. of 6 variables:..$ Ozone : int [1:153] 41 36 12 18 NA 28 23 19 8 NA.....$ Solar.R: int [1:153] 190 118 149 313 NA NA 299 99 19 194.....$ Wind : num [1:153] 7.4 8 12.6 11.5 14.3 14.9 8.6 13.8 20.1 8.6.....$ Temp : int [1:153] 67 72 74 62 56 66 65 59 61 69.....$ Month : int [1:153] 5 5 5 5 5 5 5 5 5 5.....$ Day : int [1:153] 1 2 3 4 5 6 7 8 9 10... $ layers :List of 1..$ :proto object.... $ mapping :List of 2......$ x: symbol Temp......$ y: symbol Ozone.... $ geom_params:List of 1......$ na.rm: logi FALSE. $ plot_env : - attr(*, "class")= chr "ggplot" shortened substantially Note that the internal plot specification includes the data So if you update the data, update the call to ggplot()

46 Help with learning ggplot2 Website: had.co.nz/ggplot2/ Thousands of examples! Book: ggplot2: Elegant Graphics for Data Analysis Hadley Wickham, 2009 Graphic User Interface: Deducer (R package) Ian Fellows

47 Intro to R: recap Objectives After this workshop participants will be able to: 1.Describe some capabilities and uses of R 2.Search for help within R and use good coding practices for reproducible research in R 3.Read in and summarize a simple dataset with R/JGR/Deducer Together, let’s explore some data from the WHO - Global School Health Survey. I will also give you a script containing code which you can run, modify, and take home! 4.Make some standard plots with Deducer templates

48 Open JGR -

49 Load the Deducer package

50 Note additional menus

51

52

53 Intro to R: recap Objectives After this workshop participants will be able to: 1.Describe some capabilities and uses of R 2.Search for help within R and use good coding practices for reproducible research in R 3.Read in and summarize a simple dataset with R/JGR/Deducer 4.Make some standard plots with Deducer templates Using the gshs dataframe – let's make some plots together using templates in: Deducer → Plots → Plot Builder

54 Since R, JGR, and Deducer are free, you should install them at home or work and play with them!

55 Installing R, JGR, Deducer Part I: R on Windows (shown), or Mac, or Linux R is available from a set of mirrors known as The Comprehensive R Archive Network (CRAN) http://cran.r-project.org/ Closest mirror and link for windows: http://software.rc.fas.harvard.edu/mirrors/R/bin/windows/base/ Uses a Windows installer – default options are fine

56 Installing R, JGR, Deducer Part II: JGR on Windows (shown), or Mac, or Linux JGR requires a Java Development Kit (JDK) You probably don't have this* Available free at: http://www.oracle.com/technetwork/java/javase/downloads/index.html *if you did have a JDK (and not just a JRE) you would have a folder named something like … C:\Program Files\Java\jdk1.6.0_20\

57 Installing R, JGR, Deducer Part II: JGR on Windows (shown), or Mac, or Linux JGR requires a launcher file on Windows: http://www.rforge.net/JGR/web-files/jgr-1_62.exe Leave this as your desktop shortcut

58 Installing R, JGR, Deducer Part III: Installing Deducer Deducer is an R package From within JGR To install packages: Packages & Data -> Package Installer To load packages: Packages & Data -> Package Manager

59 A few helpful R links Download R: http://cran.r-project.org/ available for Windows, Mac OS X, and Linuxhttp://cran.r-project.org/ Advice – A clearly stated question with a reproducible example is far more likely to get help. You will often find your own solution by restating where you are getting stuck in a clear and concise way. Writing reproducible examples: https://gist.github.com/270442https://gist.github.com/270442 General R links http://statmethods.net/http://statmethods.net/ Quick-R for SAS/SPSS/Stata Users - An all around excellent reference site http://www.ats.ucla.edu/stat/R/http://www.ats.ucla.edu/stat/R/Resources for learning R from UCLA with lots of examples http://www.r-bloggers.com/learning-r-for-researchers-in-psychology/http://www.r-bloggers.com/learning-r-for-researchers-in-psychology/ This is a nice listing of R resources http://stackoverflow.com/questions/tagged/rhttp://stackoverflow.com/questions/tagged/rQ&A forum for R programming questions - lots of good help! see also: http://crossvalidated.com for general stats & Rhttp://crossvalidated.com http://rstudio.org http://rstudio.org Integrated Development Environment for command line programming with R ggplot2 links http://had.co.nz/ggplot2/http://had.co.nz/ggplot2/ggplot2 help & reference – lots of examples http://groups.google.com/group/ggplot2http://groups.google.com/group/ggplot2ggplot2 user group – great for posting questions https://github.com/hadley/ggplot2/wikihttps://github.com/hadley/ggplot2/wikiggplot2 wiki: answers many FAQs, tips & tricks http://www.slideshare.net/hadley/presentations http://www.slideshare.net/hadley/presentations Over 100 presentations by Hadley Wickham, author of ggplot2. A four-part video of a ½ day workshop by him starts here: http://had.blip.tv/file/3362248/http://had.blip.tv/file/3362248/ Setting up JGR in Windows JGR requires a JDK – speak to your IT person if this seems daunting (http://www.oracle.com/technetwork/java/javase/downloads/index.html)http://www.oracle.com/technetwork/java/javase/downloads/index.html On Windows, JGR needs to be started from a launcher. For R version 2.13.0 on Windows with a 32bit R you will likely want to get the file jgr- 1_62.exe as a launcher from here: http://www.rforge.net/JGR/http://www.rforge.net/JGR/ A discussion of the features of JGR can be found in this article (starting on page 9): http://stat-computing.org/newsletter/issues/scgn-16-2.pdf Deducer - an R package which works best in a working instance of JGR – has drop-down menus for ggplot2 functionality http://www.deducer.org/pmwiki/pmwiki.php?n=Main.DeducerManual There are great videos linked here introducing the Deducer package (although the volume is quite low) This slide last updated 06/19/2011


Download ppt "Data visualization and graphic design Introducing R for data visualization Allan Just and Andrew Rundle EPIC Short Course June 21, 2011 Wickham 2008."

Similar presentations


Ads by Google