From RationalWiki
Jump to: navigation, search

DataMasher is a website that allows people to manipulate two statistical data sets, make a meaningless pretty picture, and then completely misinterpret the picture to fit their preexisting bias. It is the culmination of Mark Twain's "lies, damn lies and statistics" proverb.

DataMasher works by allowing users to select two demographic statistical sets, such as oranges per 100,000 people and apples per 100,000 people and, for a reason that is never explained, add, multiply or divide before assigning it to a color coded map of the United States. When dividing there does not seem to be any consistent method concerning the order of statistics in division, ignoring division's lack of an commutative property. This produces a list of data sets that are, statistically speaking, 93.26% irrelevant to any real world utility.

The concept[edit]

According to the website, "DataMasher helps citizens have a little fun with those data by creating mashups to visualize them in different ways and see how states compare on important issues."

In reality "have a little fun" is probably not the main usage of DataMasher. The primary usage is more likely attempts to bolster arguments on traction-free, sweaty online catfights of a political nature and likely divided over red state/blue state lines. These arguments are further encouraged by anonymous postable, largely unmoderated message boards attached to each statistical map.

Example of failure[edit]

DataMasher's Crime vs Gun Ownership is an obvious case of biased interpretation of correlated data.

The map measures "Violent crime rate per 100,000 divided by % of Households with Loaded Firearm" and many states with stricter gun laws rate on the darker, "higher" end of the spectrum while states with easy gun access rate closer to the white, "lower" end of the spectrum.

This gives the impression that more gun owners makes for less crime, and many gun nuts firearms rights proponents on the message board claimed a premature victory for the efficacy of gun owning households.

What it actually shows is that states with more guns have more guns in those states (duh).

After a look at the actual statistics, or just a thorough consideration of the two statistics being compared, the flaw becomes obvious: no matter what the violent crime rate the "Violent crime rate per 100,000 divided by % of Households with Loaded Firearm" rating can always be slid toward the lower end of the spectrum by pumping enough guns into the state, even if crime stays the same or gets worse as a result. In most cases the "low crime" rating is an indicator of the number of legally owned guns that are doing nothing at all to prevent crime in any way.

If you see a data map from DataMasher you should immediately take a skeptical stance toward it, study the original data sets and consider the relation and context of the final data. The whole site is ripe for abuse and misinterpretation.

External links[edit]