How to Lie with Statistics. Many graphs in newspapers are misleading because the authors used Excel's default graph settings, instead of making sure to set the range of the y-axis to start at zero.
Are you using a recent version of Microsoft Excel, such as Excel 2013? Excel 2013's PowerPivot and PivotChart features are amazing. [Note: I currently work as a contractor for Microsoft using these tools.] Bill Jelen (Mr. Excel) has written a book on Excel 2013 Pivot Table Data Crunching.
If you have a suitable data set (like the populations of various states over time, or baseball statistics, or...) you could have the students graph interesting data changes (like the population trend of a city or state, or a batter's on-base percentage versus time, or a batter's home-run rate with runners in scoring position, or...)
A graph of the population of a chosen city or state over time is nice. A PivotChart of the same thing is amazing. You can graph how the number of teenage girls in a city has changed over time. You can then graph how the number of 13 year-old girls and 13 year-old boys in the city has changed. Then, you can change a single filter to get a similar graph for a different city.
In my job, I try to make graphs that are very easy to understand. That usually means:
- Limiting the number of things being graphed.
- Avoiding having lots of lines cross each other in the plot area.
- A chart title that describes the point of the chart.
- Either an easy-to-read axis scale, or a data table
- Descriptive axis titles (unless the axis scale shows that the units are times)
- Consistently coloring matching things in different graphs.
- Unless I have a really good reason not to (like for date/time axes), starting axis scales at zero.
- Where practical, making the axis scales match in corresponding graphs.
- Avoiding complicated definitions for what is being graphed. It is easier to say "At-bats in May" or "At-bats before injury" than "At-bats from May 4 through June 6".
- Making sure that the field names (as shown on the chart) are easy to understand.
- Hiding redundant features (like a legend that says "Total" when there is only one variable being graphed).