Data visualization training session HELCOM
Download as PPTX, PDF2 likes400 views
The data visualization training session organized by Manuel Frias at the HELCOM Secretariat on November 28 and December 11, 2018, aimed to enhance participants' understanding of effective data visualization techniques, particularly in improving the clarity and appeal of HELCOM reports. The session covered the definitions of effective visualizations, appropriate graph types for various datasets, and design principles for graph improvement, encouraging hands-on analysis and discussions. Participants explored common visualization pitfalls and recommended resources for continued learning in the field of data visualization.
Data visualization training session HELCOM
- 1. DATA VISUALIZATION TRAINING SESSION Summary of the data visualization training session held at the HELCOM Secretariat on 28th November 2018 11th December 2018 Organized by: Manuel Frias
- 2. 2| Table of contents 3 4 6 11 12 page page page page page 13 page Whythistrainingand what’snext Part1Whatisan effectivedata visualization Part2Whatisthemost appropriategraphfora dataset Part3Howcanwe improvethestyleofa graph Threebook recommendations Sweetend
- 3. 3| Why this training and what’s next What is an effective graph and how can we improve our data visualization skills? HELCOM publishes lots of graphs in many reports every year but we admit many of them have room for improvement. Unfortunately, none of us at the HELCOM Secretariat has been taught data visualization skills. Despite this fact, we publish countless pages with innumerable graphs and visualizations. The aim of this training session was to reflect about data visualization, particularly graphs. Are there any basic rules? How can we improve our skills to make HELCOM reports clearer to understand and more attractive? This training was two hours long divided in three parts: 1.What is an effective data visualization 2.What is the most appropriate graph for a dataset 3.How can we improve the style of a graph We decided to hold informal meetings in the future to get feedback from our visualizations. Icon by Deemak Daksina TheNounProject Thepresentationused inthetrainingis on Slideshare
- 4. 4| 1 - What is an effective data visualization The session started by defining an effective data visualization. According to designer and educator Alberto Cairo, it should be: • Truthful • Functional • Beautiful • Insightful • Enlightening I showed a graph I made some years ago and we together discussed whether it met those criteria.We criticized the use of 3D, the stacked bar, the colors... All in all, it is virtually impossible to get something out of it. Icon by Arif Arif TheNounProject Book recommendations attheend
- 5. 5| We agreed that the graph had many problems and that it definitely didn’t meet Cairo’s criteria, except for being (or trying to be) truthful. I showed a remake with a simple bar chart and we all agreed that it communicated the message much better than the previous one. What was I thinking when I made the first one? Probably I didn’t think much. I just wanted to make an attractive graph without thinking in the viewer. What happened between both graphs? I just applied basic principles of data visualization. 0 0 2 2 3 3 3 3 5 5 6 8 8 9 9 Buffer Planning Clear Help Query Measure SnapShoot Download Full Extent Previous Select Zoom Out Identify Zoom In Pan Only basic functionalitiesare used in the map service Number of answers to the question "Which tool do you normally usein the map service?"
- 6. 6| 2 - What is the most appropriate graph for a dataset How do you go from the first horrendous version to the second simpler but more effective visualization? First, we should focus on finding the message. What do we want to say with the graph? For whom is it ? Do we want to emphasize anything? Do we want to show comparison, deviation...? In my first graph I skipped this step entirely. Second, we should find the shape of the data.To do this, we have two alternatives. The first one is to use a chart chooser. I recommended three. However, some have been criticized by authors like Stephen Few for offering some charts types that are not effective. I think, however, that they could be a good starter for newbies. https://datavizcatalogue.com/ https://www.data-to-viz.com/ http://ft.com/vocabulary
- 7. 7| The second option, recommended by authors like Cairo, is to use the Cleveland McGill scale.This is arguably the most famous scale in the data visualization world. It shows what kind of graphs are better to perform accurate and general comparison. Not surprisingly, the most effective graphs to show accurate comparisons are bar and line graphs. Notice that this research was done only for statistical charts. Shading and hue can be good for showing change in maps. Cleveland, McGill, 1984 adapted from Alberto Cairo, 2016: The truthful art Enable accurate comparisons Enable general comparisons Position along common but unaligned scales Length Angle Are a Volume Shading HuePosition along common scale
- 8. 8| Bar, line, point (scatterplots) and boxes charts are, in fact, the most appropriate ways to perform quantitative analyses, according to Stephen Few. Oh but wait... Where are pie charts that are used so much in HELCOM? Bad news—according to cognitive science, our brains are not very good at comparing areas, as you see in the Cleveland McGill scale. Some authors think they are excellent... for eating them! Nevertheless, I think they can be effective in some counted situations depending on your message. Source: unknown
- 9. 9| Next, we made three groups to analyze three charts that have been published in HELCOM. Our aim was to answer two questions: • Is it effective? Why? • Is there a better alternative? Sketch it After thinking together 20 minutes we came up with three new more effective visualizations. Among the solutions proposed were: avoiding clutter, choosing another type of graph and dividing the information in pieces.
- 10. 10| After the exercise, we discussed alternative ways to visualize data like, for example, slope graphs, dot plots and chord diagrams.The challenge with theses graphs is that they are not so straightforward to do in Excel 2016. It is, however, not impossible as designer Jorge Camões nicely demonstrates in his book (reference at the end).
- 11. 11| 3 - How can we improve the style of a graph In this section, we discussed how to improve the styles of the graphs we make at the Secretariat. A graph can be more visually attractive if we follow some basic design principles like, avoiding clutter, apply proper alignment, use contrast efficiently, etc. I showed a gif that become popular a while ago in social media and that summarizes nicely how to make a more effective graph in some simple steps. Finally we talked about alternative tools for exploring (Inzight) and visualizing data (Flourish and D3) Source: Darkhorse Analytics
- 12. 12| Three book recommendations There are endless books about data visualization out there. Here there are three that might interest you. Thefunctionalart byAlbertoCairo Clearintroduction Dataatwork byJorgeCamões AnythingispossibleinExcel Showmethenumbers byStephenFew (Practical)biblefordataviz
- 13. 13| Sweet end The session felt better with some buns. Unfortunately, I was late to take a a pic and these are the leftovers. It was eaten by: Alexey Bakhtov, JanaWolf, Joni Kaitaranta,Juuso Haapaniemi, Katarzyna Droździel, Owen Rowe, Sanna Saari, Manuel Sala Perez and Florent Nicolas