Paper Reading #20- The aligned rank transform for nonparametric factorial analyses using only anova procedures

Title: The aligned rank transform for nonparametric factorial analyses using only anova procedures
Reference Information:
Jacob Wobbrock. Leah Findlater, Darren Gergle, and James Higgins. "The aligned rank transform for nonparametric factorial analyses using only anova procedures". CHI '11: Proceedings of the 2011 annual conference on Human factors in computing systems. ACM, New York, NY, USA. ©2011. ISBN: 978-1-4503-0228-9.
Author Bios:
Jacob Wobbrock- He is an Associate Professor in the Information School and an Adjunct Associate Professor in the Department of Computer Science & Engineering at the University of Washington.
Leah Findlater- Is an Undergraduate Research Advisor for the Universitty of Washington. Has served on projects related to HCI.
Darren Gergle- He is an Associate Professor in the departments of Communication Studies and Electrical Engineering & Computer Science (by courtesy) at Northwestern University. He's also direct the CollabLab: The Laboratory for Collaborative Technology.
James Higgins- A professor in the department of statistics at Kansas State University.
Summary:

• Hypothesis: If the Aligned Rank Transform (ART) method can be implemented for statistical analysis and nonparametric data, then studies concerning interactions and other complex factors can be made more quantifiable and easier to manage.
• Methods: The ART method "aligns" the data before applying averaged ranks. Then, the common ANOVA procedures can be used. ARTool and ARTweb are also tools for assisting in using the ART method on relative studies. Unlike other statistical methods, ART doesn't violate ANOVA assumptions, inflate type I error rates, forego interaction effects, or disregard "too complex" of study data. the authors also made ART relatively easy to interpret- anyone familiar with ANOVA can interpret the data. The ART method can also consider "N" factors in its analysis. ART has 5 steps: 1) computing residuals 2) computing estimated effects for all main and interaction effects 3) compute aligned response Y' 4) assign averaged ranks Y'' 5) perform a full factorial ANOVA on Y''. Correctness is ensured on ART by ensuring that each column of data for Y' sums to 0 and by showing that the full-factorial ANOVA performed on Y' show all effects stripped out except for the ones which the data were aligned for initially. ARTool and ARTweb allow for simple interactions and clear data results.
• Results: Findlater collected satisfaction ratings on accuracy and interface from 24 participants. The authors found that their creation was satisfactory among both fields with the users. The authors found some limitations for ART also (extreme skew, tied ranks, not randomized designs). 
• Content: The authors wanted to create a way to handle complex data in experiments using familiar tests and data analysis already in place. ART is favored over traditional methods because of its simplicity and usability. ARTweb and ARTool allow for convenience of using ART on data and models with a computer and make ranking and alignment simple.

Discussion:
I know next to nothing about statistics. I took one course on it and forgot nearly everything relatively fast. I got lost pretty quickly in the paper, basically. It seems, however, that the field of HCI needed a unique way to interpret data in studies concerning (say) measuring if a device was "good" or not (whether users deemed it as "satisfactory" based on certain metrics). This seems to be a great idea and is also coupled with programs that allow for even easier usability with ART. This way, anyone with a basic knowledge of ANOVA and statistics can understand the ART results. I definitely believe that the authors achieved their goals because their method and programs are probably used commonly in these target types of experiments. It is good for authors to be able to see limitations on their own inventions, too. This way they can disclaimer it or they can release a "2.0" to correct some of these issues. If I needed to do some study concerning HCI interactions, I would definitely consider using ART.