Is this poll any good?
Reporters share a curiosity about their world; we’re basically information junkies. Science writers especially are driven to understand even the most complex phenomena. We think critically about the research we cover. We assess the methods of a study in order to judge whether its findings are worth sharing with the general public.
That’s why this Salon article about political polls surprised me. Despite my training as a science writer, I couldn’t assess the author’s findings. While attending an annual science writers’ meeting, I decided to take an informal poll of my own, and see if my colleagues could do a better job. Then I would consult the experts.
Ultimately, I learned something about political polls that enhanced my understanding of all the areas of science that we cover — particularly climate science.
According to the Salon author (a political pollster), if a dozen different polls, all employing different methodologies, show that a particular presidential candidate has a lead, then the lead must truly exist. As a layperson, this made sense to me. If you ask the same question many different ways and get the same answer, then it is probably right.
But is it, really?
The science writers I spoke to all seemed relieved to hear that a number of otherwise confusing polls could be summed up in one easy-to-grasp idea. Then I asked, “Do you think that logic is statistically valid?” Quizzical looks and shrugged shoulders suggested nobody knew any better than I did. Colleague Emily Caldwell reversed the question: “You cover statistics. Does it make sense to you?”
Well, I’ve covered statistics just long enough to know that there’s a lot I don’t know. But I knew who to ask.
Fritz Scheuren, vice president for statistics at the National Opinion Research Center, was on campus this week talking about political polls. His answer was similar to what he said on WOSU Radio’s “Open Line” on Oct. 29 [listen to the archive here]. To make sense of polls, people should look at how one poll from a single reputable source, such as Gallup, changes over time — not what a collection of polls from different sources says at a given moment. Different polls ask different questions of different populations in different ways, all based on the study design. The results cannot simply be lumped together. They need to be combined via a sophisticated meta-analysis.
The Salon writer’s assertion makes intuitive sense to non-statisticians, but it’s scientifically invalid.
This reminded me of some seminars I attended earlier this year which concerned how different climate change models can be combined to yield a big picture of what’s happening to the planet. Scientists are working to assemble such “climate model ensembles” in order to reduce the uncertainty that plagues such predictions. Climate models are unlike political polls, explained Ohio State statistician Noel Cressie. Climate models are closely related (they all study the same interrelated climate variables), whereas political polls should gather data from random samples of people.
But regardless, both kinds of “ensembles” need to be constructed carefully, weighing their similarities and differences. That’s the goal of meta-analysis.
To judge which political polls are reliable — in that they are good predictors of the election — Scheuren said we need to wait until after Nov. 4, and see which ones mirrored actual outcomes. But we can’t do that with climate change, and that’s why statisticians are working so hard to build scientifically valid climate model ensembles now.
The Salon article forced me to look more carefully at political polls — and at how we report on meta-analysis in climate change and other areas of science.– Pam Frost Gorder
![[del.icio.us]](http://researchnews.osu.edu/blog/wp-content/plugins/bookmarkify/delicious.png)
![[Digg]](http://researchnews.osu.edu/blog/wp-content/plugins/bookmarkify/digg.png)
![[Google]](http://researchnews.osu.edu/blog/wp-content/plugins/bookmarkify/google.png)
![[LinkedIn]](http://researchnews.osu.edu/blog/wp-content/plugins/bookmarkify/linkedin.png)
![[StumbleUpon]](http://researchnews.osu.edu/blog/wp-content/plugins/bookmarkify/stumbleupon.png)
![[Windows Live]](http://researchnews.osu.edu/blog/wp-content/plugins/bookmarkify/windowslive.png)
![[Yahoo!]](http://researchnews.osu.edu/blog/wp-content/plugins/bookmarkify/yahoo.png)
![[Email]](http://researchnews.osu.edu/blog/wp-content/plugins/bookmarkify/email.png)
Backing up, Victor Deeb is a 71-year-old professional chemist who had worked in industry for at least 20 years. Since retirement more than a decade ago, Deeb had used a makeshift chemistry lab in his basement to continue his science, most recently, he said, trying to develop a non-cancerous sealant for baby food jars.
In those days, young boys and chemistry sets seemed a perfect match. And many, like me, weren’t satisfied with the tiny bottles offered in 
It was a wonder that we survived.
But that idea has taken on a life of its own, and now there is 
But for all this travels he has one regret: He never saw the rest of those countries he visited — the people and cultures beyond the computer centers in universities and laboratories.
“Computer, show me sunrise over Mauna Kea, play Beethoven’s fifth symphony, and order that book from Amazon that I wanted… Oh, and call Mom.” And bam! There it is — whatever you want.
More than 30 countries worldwide are tied into the grid, each donating computing power to store and analyze data from the experiment. More than a dozen remote LHC computing centers around the world participated in the ceremony via videoconference, and regardless of whether the site was in Canada, Russia, China, or Australia, they all looked basically the same — a roomful of computers wired together in parallel.
As Apple’s 
The 
The particles aren’t a nuisance. Because they stream through the detector in perfectly straight lines, the scientists use them like giant rulers to gauge the alignment of detector plates. They can then program software to compensate for plates that are slightly out of alignment. __Pam Frost Gorder
Streams of protons will circulate through the smallest loop, and gradually spin faster and faster. When the beam is spinning fast enough, a series of magnetic switches diverts it to the next largest loop where the speed increases even more.
Pierre Charrue, head of CERN’s Control Infrastructure Group, said that the success of the LHC will depend on ergonomics. The scientists must be comfortable and able to focus on their work. They have to be able to read conditions in the accelerator quickly, and make good decisions. So he and his team designed the control room with a high ceiling, bright picture windows, and curved workspaces. Sophisticated soundproofing looks like textured wallpaper, in colors of blue and sand that evoke a blissful day on the beach.
Actually, CERN is a popular field trip site for school kids from all over Switzerland, France, and Germany, according to 
