Chapter 6 Audience-Focused
Just like writing, your data analysis always has an audience. Whether the audience is you, your coworker, your boss, or a policymaker, knowing your audience helps tailor how you present your findings.
6.1 Considerations for Public Sector Analysis
When sharing data analysis with a policymaker or decision maker, it is important to keep in mind that the audience will often be much wider than initially assumed. Once your analysis is in the hands of a policymaker, it may go to the press, lobbyist organizations, other interest groups, or others inside or outside your organization. Additionally, policymakers often prefer paper or other “hard copy” type analysis, limiting your flexibility to simply share a file. This poses two unique constraints:
- The mechanics of your analysis are often hidden in a print or a PDF
- Others will see your work and either take it for granted or want to dissect it.
This is where having tidy, reproducible, and flexible backup becomes incredibly important. By having this backup ready to go as soon as you present or deliver your analysis to the policymaker, you are putting yourself in a strong position to share backup when questions inevitably arise.
If you are sharing your work electronically, then there are several emerging technologies and techniques that make this process easier. It is possible to embed Excel workbooks within Word documents, which ensures that the recipient has access to both your analysis and your methods. However, this can be tricky for ensuring adequate version control and reproducible, as these workbooks will often link to data that the end user doesn’t have access to.
This problem is largely solved by using dashboards and other online solutions. Dashboards are an effective way to share analysis with decision makers because you can present your conclusions and also offer the tailored flexibility for the end user to interact as well. Tools such as PowerBI and Tableau run on the same tidy data and pivot-oriented platform discussed here. It is also straightforward to share the mechanics of the model with anyone who is interested– especially if you follow the best practices for reproducibility outlined earlier. The downside of such dashboard solutions is they can be expensive, proprietary, and it can be difficult to configure them for specific audiences.
6.2 An Introduction to Literate Data Analysis
That brings me to what form democratic data analysis would take in the World-According-To-Corban. Another tool for presenting data analysis to decision-makers are interactive notebooks that merge explanatory text, data, code, and graphics in one file that can be tailored for a variety of audiences. I call this approach literate data analysis7. Code-and-text driven notebooks are becoming exceedingly common in academic domains. In fact, this handbook is an example of an interactive HTML document written in r8, but they certainly do not have to be this involved. Interactive notebooks combine text, code, and output in one place. They can be structured so that the file shows either plain-text analysis and charts in a web browser, but all of the code that generates it if you open the file in an editor such as RStudio. For example, here is an R notebook that I authored for a research project on legislative voting behavior, and I’ll link to other interactive notebooks in the Practices section.
Here are three platforms of interactive, code-driven data analysis notebooks:
I firmly believe that the notebook approach is the future of policy and data analysis in the public sector. These tools are free, robust, and accessible to those with varying levels of technical expertise. They also inherently solve the challenges of reproducible data analysis that I have been examining here. Version-controlled interactive notebooks are flexible, shareable, editable, and allow both decision-makers and analysts to use the same file to either draw conclusions or validate results.
I dream of a future when budgetary fiscal notes, quantitative bill analysis, and data models in the public sector are written as interactive notebooks, with explanatory text and analysis bundled together, version controlled, and presented with assumptions clearly documented.
This isn’t to say that there aren’t downsides to these approaches. The largest, as it currently stands, is the learning curve required to use these tools. I have yet to come across a notebook format that is as easily intuitive as an Excel workbook9. For now, the biggest downside of these tools is that they require at least a baseline understanding of r, python, javascript, or another similar scripting language. As we have already covered, Excel is a programming language, and probably the world’s most common one at that. But Excel is familiar and relatively beginner friendly because it hides the fact that the user is actually programming.
Once that baseline level of understanding is acquired, the notebook format becomes an intuitive output for reports. The same document can result in a PDF file to be shared with decision makers, an interactive document to be shared online, and a complete record of your analysis for a peer. Long story short, I’m excited to work towards integrating the notebook philosophy to the public sector as the next frontier of democratic data analysis.
Yihui Xie, the creator of Rmarkdown, summarizes these principles nicely: > I think notebooks are popular for the same reason that explains the popularity of spreadsheets such as Excel. I haven’t met a single software engineer who loves Excel. Everyone hates it and makes fun of it, but why do so many users still use it? Again, Excel makes things tangible. You can touch the data (although it is usually a very bad idea), and draw graphics in a sheet that contains the source data (bad idea again). It makes you feel everything is well under your control: oh here is my data, and here is a graph next to it; oh I should use that column to draw the graph instead, so let me change it and I can see the updated graph immediately.2 You can do everything in a single place, and the short distance between the source (data) and the output is ace.
Excel makes things tangible at the price of making things messy (e.g., it may contain manually edited data that is hard to keep track of, or merged cells or graphs that make it hard for other software to read the data). By comparison, although notebooks can mess up the state, but that is only an intermediate problem. At its core, it is still relatively clean and encourages the reproducibility principle, i.e., you shall use code to generate results automatically instead of manually copying and pasting results in your report. If you are concerned about the state, you can restart the session and recompile the whole notebook from scratch. Spreadsheets are often hopeless here—you cannot easily restart your brain and redo exactly the same things.
-Yihui Xie, Creator of Rmarkdown on notebooks and excel