Statistics Professor Hides Pictures, Messages in Problem Solutions

Say you’re an aspiring statistician who has just spent hours trying to figure out the answer to a particularly thorny problem. As you plug the final numbers into the computer program you’re running in order to confirm your analysis, an image takes shape on the screen: Homer Simpson, working a problem on a blackboard. What would your reaction be?

For students of Dr. Len Stefanski, professor of statistics at North Carolina State University, it’s usually more “Whoa!” than “D’oh!”

Stefanski has developed a method for hiding messages and pictures in data sets that only appear when a student or statistician has correctly analyzed the data via regression analysis, a statistical process for studying how one variable depends on others.

His innovative approach to making statistical analysis fun is featured in the May edition of The American Statistician.

“Regression analysis is used in every scientific field,” Stefanski says. “It’s a way to describe how one variable depends upon other variables – for example, you can use regression analysis to discover how blood pressure is affected by cholesterol, weight, diet and other risk factors. Or how temperatures in different parts of the ocean might affect the number of hurricanes that occur in a given season.”

When statisticians do regression analysis, they use computer programs to help them discover trends and variability within a given set of data. The programs plot the data as dots in an onscreen graph. A statistician looking at the relationship between height and weight, for example, would expect the general trend in the graph to be that weight increases with height – and that the majority of the dots in the graph would fall along that trend. However, there are always exceptions – people who don’t fit the pattern or trend. These exceptions would show up as “scatter” in the computer model, or random dots on the screen.

“The challenge for the statistician is to find the correct trend in a data set,” Stefanski says. “If you’ve done the analysis correctly and extracted the trend from the data, then all you should be left with onscreen when you’re finished is the random scatter, and I wanted to find a way to make the payoff for getting the right answer a bit more interesting for students.”

Stefanski created a simple computer program that allowed him to “hide” images or messages in data sets. When a student successfully identifies and removes the trend data from the set, the message or picture is revealed in what is known as a residual plot. He has made the data sets available to colleagues and the public on his Web site, and will make the computer program he created available soon as well.

“It’s not a terribly efficient means of encryption,” Stefanski says, “but it certainly makes statistical analysis more visually interesting.”

Source: NC State University

Related stories:

Researchers develop next-generation computer antivirus system
(PhysOrg.com) -- Antivirus software on your personal computer could become a thing of the past thanks to a new "cloud computing" approach to malicious software detection developed at the University of Michigan. Cloud computing refers to applications and services provided seamlessly on the Internet.
Researchers devise new method for protecting private data
Companies and organizations that keep sensitive personal information on millions of Americans have become attractive targets for hackers in recent years, resulting in billions of dollars in losses for U.S. businesses and misery for countless consumers.
IBM researchers quell nanoscale interference
IBM researchers have discovered a way to use graphite effectively in building nanoelectonic circuits vastly smaller than those in silicon-based computer chips.
Tooling up for tomorrow's clever cars
Cars are becoming more complex, with a range of advanced features we could hardly have imagined a few years ago made possible by sophisticated software-driven electronics. The downside is, with more to go wrong, more is going wrong, but European researchers have developed an antidote: a new computer language.
Hammer Storage Pounds Out 'Disruptive' 1TB Appliance
The new device resembles a microwave oven, offers remote access and provides the functionality of a network storage device at a price that may boggle some minds.
A window on the environment from Tahoe to the ocean
If there is a big idea in environmental science, it is interconnectedness. Air, water, land and the living things on or in them are connected in ways obvious and subtle. But actually seeing and understanding all those connections in context is very hard. That is why environmental and computer scientists at UC Davis are working together to get a "God's-eye view" of a chunk of California, from the ocean off Bodega Bay to Lake Tahoe.
Parts of Caribbean and Central America Likely to Have Less Summer Rain
Parts of the Caribbean and Central America are likely to experience a significant summer drying trend by the middle of this century, UCLA atmospheric scientists will report in the April 18 issue of Proceedings of the National Academy of Sciences (PNAS).
Simulations shed light on Earth's history of magnetic field reversals
A new analysis of computer simulations of Earth's magnetic field suggests that its behavior was different early in Earth's history, resulting in greater stability and fewer reversals of the magnetic field. The findings by researchers at the University of California, Santa Cruz, are helping to reconcile the geologic record of magnetic field reversals with the current understanding of how the Earth's core generates the planet's magnetic field.

News discussion:

General Science news