How Not to Present Data
I have a confession to make. I have a deep and abiding fondness for data visualization. I think it helps epidemiologists, public health workers and scientists in general understand things in ways numbers, however compelling, cannot. But like statistics, graphs can be cooked, manipulated and generally wiggled around until the information they convey is lost.
Some of the best examples of this are pulled from outside science, or at least outside the health sciences. Over at Economist’s View, awhile back, there was a pretty well thought out gutting of an example of the Laffer Curve using ‘real world’ data and a enormously faulty graph.
But today’s comes from Fortune. It was brought to my attention by my lovely partner in crime, who might have been distracting herself in class by looking into why Apple’s share price went tumbling the last few days (more-so than warranted perhaps by general market tumbling). Disclaimer: I do own stock in Apple, although not much, and really, I don’t expect this blog to boost its share price.
Morgan Stanley’s Kathryn Huberty and Mike Abramsky, a capital analyst for RBC both downgraded the stock recently. Abramsky cited a survey they did with ChangeWave showing that the planned purchases of Apple computers by corporations may have peaked, and were headed downward. The graph in question:
A fine graph. The problem here is not the actual graph – the graph is fine, conveys what it needs to convey, etc. The problem is results without context. A single graph, like a single statistical point estimate, is often meaningless if its not in context. The two graphs accompanying the Apple graph above:
Now we start getting a different picture. Yes, planned purchases of Apple computers have taken a dip. But planned purchases of other computer brands have been taking a dip for awhile now. Whether or not the analysis of Apple’s stock price and future is wrong or not, it does provide an interesting real world example of presenting a single graph without the underlying context, and how the interpretation of the data can change based on it.
Filed under: General, Macs in Research | Leave a Comment