My educational background is in statistics, psychology, and computer science. I’m sometimes skeptical about all three. A good case in point is various analyses of the results of the 2016 U.S. presidential election.
Here’s a map of election results generated by a researcher at the University of Michigan. It shows results by county where red indicates the county voted for Trump and Blue indicates Clinton. The map suggests a landslide win for Trump.
Now here’s a bar graph that I made which shows results by popular vote. The best information I could find stated that Clinton won 65,844,610 votes and Trump won 62,979636 votes. When I placed the data in Excel and told Excel to make a graph, this is the default graph that was generated:
The graph suggests a landslide win for Clinton. But the graph starts the y-axis at 61,500,000 votes. So I told Excel to start the y-axis at 0 and got this graph:
The graph now suggests a virtual tie.
Well, the point is that statistics can be made to tell very different stories.