Data visualizations are the modern equivalent of visual communication with the goal of better communicating numerical information to an audience. Interactive visualizations allow the audience to explore and engage with the data that is of particular interest to them. Effective visualizations make complex data more accessible, understandable, and usable.
Visualizations take many shapes and forms, and like any form of communication, a great deal of artful creativity accompanies whatever scientific method in devising the best visualizations. That said, however, a few key principles go a long way in thinking about visualizing data.
First, visualization is all about patterns, about categorizing or aggregating data in meaningful ways, and then revealing these patterns in ways that require the least amount of brain power. And so, when looking at our raw data, we need to identify which fields might serve as categories or dimensions, and which fields might serve as numeric values and how those numeric values might be aggregated in meaningful ways.
Second, visualization is about communicating a message to an audience. Visualization software allow us an enormous variety of visualization options to choose from. However, in some cases one simple number may be the most effective means to communicate a results. What decides this is knowing what exactly it is you want to communicate with your product, i.e. your message, and to whom you are communicating, i.e., your audience.
Finally, experience is key to getting better at constructing visualizations. Again, like writing or other forms of communication, you get better the more you do. Take a moment to reflect on what you like or don't like about other visualizations you see. Also, seek other's opinion about ones you create. Develop your own common sense about what makes a good visualization!
There are a plethora of tools to help researchers display their data to a broader audience. Some of these tools are:
Shiny R: https://shiny.rstudio.com/
Python: many libraries (such as Bokeh and pandas) can be used to create interactive data visualizations. Python also has tools such as:
Tethys Platform: http://www.tethysplatform.org/
Tableau: https://www.tableau.com/
In the interest of time, we are going to explore Tableau. Unlike R or Python based visualization tools, Tableau is proprietary software. As a student, you can string along as many free 1-year licenses as you want (as long as you remain a student), but if you want to continue using Tableau beyond academia, you'll need to pay for it. However, Tableau's learning curve is much shorter, allowing us to dive deeper into aspects of data visualization.
Tableau is not installed on NSOE machines. Instead, you'll need to install your own copy. You can either apply for a 1-year license of the full desktop edition or download the free public version here: https://public.tableau.com/s/download/public/pc64 . The key difference between the two is that the public version does not allow you to save your Tableau workbooks...
In our case, we' will be installing Tableau on virtual machines. Instructions for that are provided here.
We begin with a crash course on Tableau, using it to import, explore, and visualize water use data compiled from the USGS. Learning objectives include:
Importing data into Tableau
Manipulating and organizing data
Plotting in Tableau: Bar plots, horizontal bar plots, pie charts, & choropleth maps
Creating dashboards and storyboards
Uploading your results to Tableau's server for sharing
ggplot within Pythonpandas andmatplotlib for quick plotsseaborn for more complex plots: countplots and heatmapsfolium for map visualizationsggplot within R