STATISTICA Blogs

One thing you have to do when making graphs is to be sure that your graph actually makes things easier for your reader. If it is too convoluted or “scary” looking, it turns your reader off and makes them wish you just gave them a list of written facts or a column of numbers instead.

I have a tale about a messy graph. It is fresh on my mind.

First, some background on this graph: The monthly job report figures are out today and, as always, there is speculation as to what, if any, impact they will have on the upcoming Presidential election. I wanted to see what has happened historically to see whether the Obama Campaign should be worried. I grabbed data from the Bureau of Labor Statistics going back to 1939. It captures the percentage of jobs gained or lost in the twelve months since June of the previous year. I chose June because it allowed me to use the most recent data possible.

That left me with two columns of data – the year and the percentage change. To me, this seems like the perfect use for a Line Plot. It allows you to order your points sequentially from left to right and to connect them so that your reader can see the patterns more easily.

Once I had my line plot made, I saw the aforementioned “scary” graph. In short, it needed cleaning up more than my kitchen does.

I considered removing the first several points. They had high values that increased the range of the Y axis all the way to 16. Despite my earlier writing about the ethics of changing the scaling of a graph, I didn’t see how removing the points led the rest of the graph to be misleading so I decided to proceed in order to enhance readability of the graph.

Include/exclude data as appropriate: If you double click on the graph, you will bring up the Graph Options dialog, which allows you to customize a graph you have already created through the analysis dialogs. On the Graph -> Layout panel, click the Details button, and then click Select Conditions. This allows you to include or exclude specific rows of data without having to go back and re-create your graph.

I excluded the first four points and returned to my graph to see what was next. 

Label only the points you care about: I used the Brushing tool to click on the points that I wanted to display a label for in the graph. The points I chose were years when a sitting president (incumbent) was running for reelection. I labeled them with the name of the president and whether they were reelected or defeated.

But my graph was still too cluttered. 

Customize your labels:  I used the mouse to move the point labels around so that they weren’t blocking other labels, which helped a lot, but I knew it could be better. I had the text of “reelected” or “defeated” on each of the labels. Why clutter the graph with words when I can use intuitive colors to represent defeat? 

There is a tab for Point Labels on the Graph Options dialog, and a way to customize the text for each label. I changed the names of the Presidents who were defeated in their reelection campaigns to have a red text color.  

I know I feel defeated when I pull up to a red light or get red marks on a test, so that seemed like a natural choice. I added a subtitle to the graph to explain my color scheme.

Now we can see what’s going on.  It’s clear from the text color that three presidents out of ten have been defeated in their reelection campaigns since 1948. 

What patterns are there?

Display Y axis coordinates and Reference Lines:

I turned on the Y coordinates and created a dark reference line at 0 when I started thinking about summarizing this graph. I wanted to be able to say things like “Despite 3.7% job growth from June of 1975 to June of 1976, Ford was defeated” and “Despite two consecutive years of negative job growth during his first term, George W. Bush was reelected in 2004.”

After we customized the graph to make it cleaner, I can now see that there is definitely not a clear pattern here about the relationship between job growth and Presidential reelection. And it was far easier and less smelly than cleaning my kitchen. 

Image Credit: http://www.flickr.com/photos/jurvetson/1341978643/