How to Lie with Statistics

Posted on

When looking at a chart, a sharper increase in slope automatically assumes more progress in whatever the graph is showing to a reader. The extreme increase connotes more progress over time because it appears to be climbing more rapidly; depending on the scale, this may not be the case. Graph A begins higher up than graph B, which also contributes to why it would seem like it is making more progress. It seems as though it is not starting from the bottom of the graph and therefore graph A is positioned higher. I created the two charts and altered their positioning by changing the maximums and minimums. This changed the slope of the graphs, while keeping the data points the same. They are portraying the same information, but in different ways that can sway a reader or a particular argument one way or another. In graph A the maximum was .18 while the minimum was 0. For graph B the minimum was also 0 but the maximum was 1.0. The increased maximum from A to B is what altered the slope of the graph and also changes how the graph is read. 

Charts–”How to Lie with Statistics”

Posted on

To lie about the statistics, I altered the minimum and maximum of the Y-axis to portray the data in a way that would make it appear drastically different from what a simple data table—or an appropriately scaled graph—would portray to the viewer. By increasing the maximum of the Y-axis, the graph made the results of the data appear as though the actual and legal progress towards reaching the 30% proposed goal of the Sheff 1 case (2003-2007) was not only reached—but fairly insignificant of a progress (as you can see, the lines appear as though they are essentially one). On the other hand, increasing the minimum and decreasing the maximum on the scale of the Y-axis made it appear to the viewer as though not only was there little progress made over the years—however, the actual proposed goal of 30% seems blatantly unachievable in the near future.

Not Altered–"Original" Graph

 

This exercise proved itself rather fascinating and showed how fairly detrimental and deceiving visuals can be in respect to portraying data. What I mean by this is that human beings are very visual and tend to look at the shape of the graph and not really read into the actual numbers. Although all of the three graphs present the same data and same numbers, they are portraying completely different results visually.

To create the graph where it appears as though virtually no change or progress was made towards reaching the 30% integration goal of schools by 2007, I changed the minimum of the Y-axis to 14 and the maximum of the Y-axis to 30. I also changed the major unit to 0.8 and the minor unit to 0.1. By increasing the minimum, It made the line appear as though very insignificant growth on the “progression department” had occurred. By decreasing the maximum to the actual goal of 30, it made the other line appear at the very top. The major and minor units changed the intervals between by increasing them—furthering the distance between the actual progress of the state and the proposed goal of the original Sheff I hearing.

Altered Graph–Data Appears As Though No Progress Was Made Towards Obtaining Goal

In respect to the graph where it appears as though there was great progress and that the proposed goal of 30% was met, I decreased the minimum of the Y-axis to -50 and increased the maximum of the Y-axis to 2000. Additionally, I played around a bit and changed the major unit to 200 and the minor unit to 50. By making all of the alterations listed, it looks as though not only did the state meet the proposed goal—but it wasn’t far below the goal of 30% to begin with—virtually the two lines appear as though they intersect. Clearly, this is not the case, because Sheff II and a newly negotiated proposed goal came into effect shortly after the state failed to meet the original goal of Sheff I.

Altered Graph–Progress Was Made

 

How to Lie with Statistics

Posted on

To create the following two graphs, I first extracted data from the “Actual and Legal Progress toward Sheff I Goal, 2003-2007″ chart featured as Figure 5.1 in Dougherty’s Sheff v. O’Neill: Weak Desegration Remedies and Strong Disincentives in Connecticut, 1996-2008. In an excel spreadsheet, I entered years into Row A and the corresponding progress percentages into Row B. I then highlighted the data, accessed sub menus at the top of the excel menu bar, and selected a line graph. To skew perspectives of progression, I altered the scale of each graph by adjusting the axis limits to portray extremes.

The first graph, pictured to the right, denotes a drastic positive change in over just a few short years. To achieve this look I set the axis limit to 0.32, just under the 0.35 percent mark, to represent a skewed data plot illustrating sweeping positive progression.
In the second graph, I strategically selected axis information that would cause the reader to speculate a less dramatic change occurred over the same short time period. To produce this graph, I increased the range of the axis significantly and even put the range into negative figures causing the reader to infer that growth was steady but slow.
These two different depictions of data represent the same information in the same format with the goal of eliciting a certain response from its audience. Altering just a few seemingly insignificant settings yields significantly different outcomes. The power of this post is revealed in the simple truth that anyone, anywhere can alter any set of data to support virtually any argument by portraying a certain set of facts (that unfortunately an overwhelming majority of readers assume to be concrete, fixed, and true) in a mischievous light. Ultimately, it is important to recognize that statistics don’t always tell a true story.