Beautiful Data Should be Tested Before You Present

Written by Dan Gordon Wednesday, 18 May 2011 06:20

Infographics can be favorite posts for bloggers and readers alike. Without a doubt you've seen them on top websites including Mashable, TechCrunch, and Information is Beautiful. Perhaps you might have also noticed them reblogged all over Tumblr as well as retweeted via Twitter. These amazing, aesthetic representations of information earn considerably more interest and go viral a good deal sooner than the same information written down. At the same time, the hazards are not as understandable and many top ranking websites post stunning infographics that offer little insight. When you start creating a visual representation, ask yourself first: is it significant?

Infographics can be favorite posts for bloggers and readers alike. Without a doubt you've seen them on top websites including Mashable, TechCrunch, and Information is Beautiful. Perhaps you might have also noticed them reblogged all over Tumblr as well as retweeted via Twitter. These amazing, aesthetic representations of information earn considerably more interest and go viral a good deal sooner than the same information written down. At the same time, the hazards are not as understandable and many top ranking websites post stunning infographics that offer little insight. When you start creating a visual representation, ask yourself first: is it significant?

To market their own research arm LinkedInsights, Linkedin produced a stunning infographic listing the top 3 names for CEOs of both genders. The trouble? Even if there are more CEOs named 'Jack' doesn't prove that this information is significant at all. Consider the following situation: we have one hundred CEOs in room, three are named 'Peter', 2 named 'Jack', two named 'Bob', the other ninty-three CEOs all have unique, unmatching names. Once we collected this data we could confidently claim that 'Jack' was the most popular name among CEOs ... but so what?

Just because a trend is observable doesn't mean it offers any significant insight. Test for statistical significance before you start producing that infographic

How do you do this? You will need a statistical software package like SPSS produced by IBM or even the no cost PSPP. Both will offer several testing techniques when comparing and processing your data. Take care to decide on a test that matches the data you have.

Will you be comparing means or several components? Are your factors dependent? How convinced do you think you are? Better figures are often had by accepting lower than 95% confidence rate, though ultimately this does your study a disservice.