Conclusion

Next steps

This book has been, of necessity, a short one that was designed primarily to give you a flavor what R could do for you and how you could use it in your own work. The depth and breadth of R—especially when coupled with its contributed packages—make it impossible for any one book to do it justice. But hopefully this will serve as a jumping off point for some of your own explorations with R.

Before we go, I wanted to give you a short list of possible next steps for learning about statistical programming and data analysis. These resources include books and websites, associated software, and local and global events for users of R.

Books

The growing popularity of R is mirrored in the increasing collection of books that are available. Some excellent resources include:

• An Introduction to R by The R Core Team, available here.
• The R Book by Michael J. Crawley
• Statistics: An Introduction Using R by Michael J. Crawley
• R in a Nutshell: A Desktop Quick Reference (2e) by Joseph Adler
• R Cookbook by Paul Teetor
• R Graphics Cookbook by Winston Chang
• ggplot2: Elegant Graphics for Data Analysis by Hadley Wickham
• Lattice: Multivariate Data Visualization with R by Deepayan Sarkar
• Learning RStudio for R Statistical Computing by Mark van der Loo and Edwin de Jonge
• Getting Started with RStudio by John Verzani
• Many more are listed here.

Websites

In addition to published books, several online resources are of great value. Surprisingly, a conventional Google search with the letter R and the topic of interest will nearly always lead to sites related to this software. Here are a few resources I’ve found:

• The R Project website
• The R Journal is the open access, refereed journal of the R project for statistical computing.
• R-bloggers, news and tutorials about R, contributed by over 400 bloggers. There are 200-300 new posts each month.
• Rseek.org by Sasha Goodman
• Stack Overflow has great discussions on R.
• Wikibooks: R Programming Wikibook
• Quick-R

• ggplot2 website by Hadley Wickham

Software

For developers who are comfortable working with C++, Dirk Eddelbuettel and Romain Francois have developed Rcpp, an integration of R and C++ that can provide dramatic improvements in processing speed.

• Rcpp by Dirk Eddelbuettel and Romain Francois

• Tutorial by Hadley Wickham

Events

International and local groups meet to present research using R and new methods that can be applied in your own research:

• useR! is an international conference that takes place in June or July of each year

• The 2013 conference was at the University of Castilla-La Mancha in Albacete, Spain.
• The 2014 conference was at UCLA in Los Angeles, California.

• Local R User Groups

• See the full list at Revolution Analytics.

Conclusion

Hopefully this book has left you excited about the possibilities of using R, and motivated to learn more and even return the favor by contributing to the R community. Thanks for joining me. Good luck with your own explorations!

[2] In the current version of R—I am using version 3.0.3 as I write this—it is not always necessary to put quotes around the package name. I would still recommend that you use quotes around the package names for two reasons: (1) it increases cross-version compatibility, and (2) this is how the code appears in the console if you check the package by hand in RStudio’s package list.

[3] It is possible to run detach() without quotes around the package name, like this: detach(package:ggplot2). It is also possible to run the command without specifying “unload = TRUE,” but you could have problems with uncleared namespaces. The reason I suggest detach("package:ggplot2", unload = TRUE) is because that is how R shows the code when you uncheck the package by hand. This is the method that is least prone to errors. Also, if you receive an invalid 'name' argument error, then add character.only = TRUE to the command.

[4] See http://cran.r-project.org/doc/manuals/R-data.html#Reading-Excel-spreadsheets

[5] To save an SPSS SAV file as a CSV file, use these two options in the “Save As” dialog: (a) “Write variable names to spreadsheet”; and (b) “Save value labels where defined instead of data values.”

[6] See http://vudlab.com/simpsons/, which has insightful commentary and interactive graphics for this data set.

[7] Another great resource for color palettes is Adobe’s Kuler website at https://kuler.adobe.com.

[8] See http://en.wikipedia.org/wiki/Level_of_measurement for more information.

[10] If categories are combined to match R’s four categories: Irises with specks and dark brown irises are combined with brown; blue/grey irises are combined with blue; blue/grey/green irises with brown/yellow specks are combined with hazel, and green/light brown irises with minimal specks are combined with green.

[11] The confidence interval for the median is calculated as +/-1.58 IQR/sqrt(n). See ?boxplot.stats for more information.

[12] Strangely, R labels this island as “Asia,” which is not correct. It should say “New Guinea.” The other labels are correct. The reason for the switch is not clear.

[13] There is, in fact, a theoretical justification for deleting the highest seven outliers from the original data set. This justification will be discussed in the section on dichotomization.

[14] Another code is available, NaN for “not a number,” but that doesn’t usually apply to most missing data situations. For more information on both codes, see ?NA and ?NaN, respectively.

[15] See http://en.wikipedia.org/wiki/De_Tomaso_Pantera

[16] See ?iris or http://en.wikipedia.org/wiki/Iris_flower_data_set for more information.

[17] This package includes functions and data sets that are included in Modern Applied Statistics with S, a 2002 textbook by Venables and Ripley. See help(package = "MASS") for more information.

[18] For an interesting follow-up on this data, see "Taste Endures! The Rankings of Roger de Piles (†1709) and Three Centuries of Art Prices" by Kathryn Graddy at http://people.brandeis.edu/~kgraddy/published%20papers/DePiles_complete.pdf.

[19] See ?TukeyHSD for more information on this procedure in particular. See ?pairwise.t.test and ?p.adjust for more information on post-hoc tests in R and specific procedures available.

[20] Out of curiosity, I generated new data and ran the process a few more times. The second time, group 2 was different from groups 3 and 4 but that was all. The third time, group 4 was different from 1 and 2 and that was all. Again, the lesson is that larger samples are required for more stable estimates.

[21] See the Regression Modeling Strategies package rms for one excellent example.

[22] That line is from page 185 of Norman Cliff’s 1987 book, Analyzing Multivariate Data. Similar sentiments are expressed by Bruce Thompson in his 1998 talk “Five Methodology Errors in Educational Research: The Pantheon of Statistical Significance and Other Faux Pas,” which lists stepwise methods as Error #1 (see http://people.cehd.tamu.edu/~bthompson/aeraaddr.htm), or his 1989 journal editorial entitled “Why won't stepwise methods die?” in Measurement and Evaluation in Counseling and Development (see http://myweb.brooklyn.liu.edu/cortiz/PDF%20Files/Why%20Wont%20Stepwise%20Methods%20Die.pdf). In simpler terms: analysts beware.

[23] See page 230 of http://cran.r-project.org/web/packages/psych/psych.pdf.