Misleading Graphs

Author by Megan Dehn

Picking the right Visuals for your Data

At Concurrency, employees have access to many graphical advantages through our company’s partnership with Microsoft tools, like Power BI. From bar charts, to pie charts, waterfall charts, tree maps, area maps, and donut charts (yummy!), we have been introduced to the very colorful world of Power BI visuals. Understanding how to accurately use these visuals without creating misleading graphs is what’s most important, though. A misleading graph misrepresents data, creating an incorrect conclusion derived from the graph. These graphs may be intentional but can also be accidentally poorly constructed.

 

Misleading graphs: How do you know?

Some graphs are intended to mislead, and some are intended to shock. If you don’t want to do this then you need to be honest with your data. Do not implement biased opinions into your work, in hopes of persuading a client in your direction. Power BI tools are helpful because all the immediate visualizations are constructed in a way that is not misleading.

 

MANIPULATING THE Y-AXIS

I have created two visuals for my fellow readers. Each figure uses the same exact data, comparing the same amount of sales dispersed over four years, but they appear to be two extremely different graphs. Can you guess which graph is “misleading” to the general audience?

Sales-by-year-(1).JPG

salesbyyear2-(3).JPG

I’ll give you a hint.

It’s Figure 1! Congratulations! Find me, and you can get a high five.

Figure 1 is a misleading graph because the y-axis starts at $160,000 in sales and not zero. This allows the sales in 2018 to appear much more than the sales in 2017. When compared to all four years, the sales in 2017 we’re not drastically lower than the sales in 2018. In Figure 2, the scale is proportional to the data, which accurately shows the distribution of sales over the course of four years.

Hopefully, with the help of the following tips, you can make sure to look out for misleading graphs the next time you’re working in Power BI or any other data visualization software.

  1. Determine your visualization goal
    1. Inform: give out a single important message or data point that doesn't require much context to understand
    2. Compare: show similarities or differences between data series or compare amount values or parts of a whole
    3. Change: visualize trends over time or space (location)
    4. Organize: show groups, patterns, rank or order
    5. Reveal relationships: show correlations among variables or values
  2. Choose the best chart to achieve that goal
    1. Best charts for informing
      1. Use big, bold, colorful text
        1. In Power BI, you can use installed graphical features such as a “Table”, “Card”, “KPI”, “Multi-row Card”, “Donut Chart”
    2. Best charts for comparing
      1. Bar or column chart to compare independent values
      2. Use a bubble chart to compare independent values with clear outliers
        1. Explicitly label each bubble with its value
      3. Use a pie chart to compare parts of a whole
        1. Limit the chart to a maximum of 7 segments
      4. Use a stacked bar or stacked column chart to compare the compositions of multiple values
    3. Best charts for showing change
      1. Use a line chart or an area chart to show changes that are continuous over time
        1. Use area charts for less than 4 data series
      2. Timeline infographics to visualize events
      3. Use a map series to show changes in location data over time
        1. Use a choropleth map to show location data
    4. Best charts for organizing
      1. Use a numbered list to show rank or order
      2. Use a table for readers to look up values
      3. Use boxes borders, arrows, and lines to visually organize groups
    5. Best charts far revealing relationships
      1. Use a scatter plot to reveal the correlation and distribution of a two-variable dataset
      2. Use a histogram to reveal the distribution of a single variable