Trend That Is Not A Trend: Creating a Trend From Measurement Changes
I was watching some excellent videos of recent Phoenix dust storms roll across the city. I started thinking about a joke story:
Scientists report that the number of Phoenix dust storms have likely increased substantially since 1990. Before that date, almost no cell phone videos exist of large dust storms in Phoenix. Today, one can find hundreds of such videos on Youtube, mostly from the last three or four years. Obviously we are seeing some sort of climate change
This would clearly be absurd -- there has been a change in measurement technology. No cell phone cameras existed before 1990. But equally absurd examples can be found every day.
- With the summer of the shark, an increase in frequency of media coverage of shark attacks was mistaken for an increase in frequency of shark attacks themselves.
- With tornadoes, improving detection of smaller twisters (e.g. by doppler radar and storm chasers) has been mistaken by many (cough Al Gore cough) for an increase in the frequency of tornadoes. In fact, all evidence points to declining tornado frequency
- With electrical grid disturbances, a trend was created solely by the government owner of the data making a push with power companies to provide more complete reporting.
- I have wondered whether the so-called cancer epidemic in India is real, or the results of better diagnosis and longer life spans
Postscript: I remember when I first saw one of these storms rolling towards me after I moved to Phoenix. Perhaps I should not have read Stephen King's The Mist, but I honestly wondered for a minute if I would live to regret not hopping in my car and racing to stay ahead of the wall coming towards me.
"Trend that is not a trend" is an occasional feature on this blog. I could probably write three stories a day on this topic if I wished. The media is filled with stories of supposed trends based on single data points or anecdotes rather than, you know, actual trend data. More stories of this type are here. It is not unusual to find that the trend data often support a trend in the opposite direction as claimed by media articles.
Craig Loehle:
There are plenty of other examples. It is claimed that gun violence is rising, when it has been falling for 20 yrs based on FBI statistics but our response to mass shootings has gone up. Likewise, it is claimed that there is an "epidemic" of sexual assault on campus but these statistics are also down. People claim that cancer is on the rise when deaths are falling.
July 11, 2014, 7:11 amRick Summerson:
Yes but the news would be much, much duller if it was confined to facts. Just think how much more boring the world is to me because I had to take statistics classes in school.
July 11, 2014, 7:42 ammahtso:
So cell-phone videos cause dust storms. What else could it be?
July 11, 2014, 8:00 amFelineCannonball:
Those fake scientists looking at YouTube video frequency in your joke story sure are stupid. Ha ha ha.
This is what ice cores, lake sediment, and sediment traps are for. There are great regional dust records. Each has to be interpreted carefully but there are some obvious things that come out on a variety of scales. Arizona dust storms are recorded in Colorado lakes and snowpack and ice. Glacial-interglacial shifts are giant, and its clear that year over year differences in dust can shift snowmelt by a month or more. On the long time scale glacial dust is a good example of a strong negative feedback in ice sheet dynamics. They create dust that lowers their albedo and limits their extents. Soot and volcanic dust also have significant impacts -- something that shows up in quite a few regional stories. Give it another 500 years and the anthropogenic story will be obvious, whatever it is. My understanding is that there are competing processes driving dust volume, but if you care about colorado river basin snow pack you should pay attention to winter and spring dust storms and forget about summer haboobs.
You got me with epidemiologists and the media though.
July 11, 2014, 8:29 ammarque2:
Detection vs correlation is a real problem for statistics. For an instance we hear Asthma is on the increase - but it turns out the increase is mostly due to awareness comapaigns that got doctors and parents worked up increasing the detection rates. Great that more people are or being helped - but the trend isn't upward due to pollution or CO2 as some folks claim.
Warren is right about tornadoes as well - we now have better ability to detect them. Hurricanes too. You used to have to fly an airplane into the vortex of a topical storm and if you were lucky on marginal ones you would see the correct barometric pressure. You would only fly once or twice a day. Now with satellites you can see the barometric pressure and detect the exact minute it became classified hurricane - this cause upward bias on the low level as well.
July 11, 2014, 8:54 amFelineCannonball:
Epidemiology, or rather media depictions of epidemiology results has real problems. There are obvious publication biases against null results, correlation/causation problems, missing variables, bad collection procedures, mixed data sets, changing external variables in long term studies, etc. People in the field know this and take everything with a grain of salt -- looking at the details , but the rest of should probably just stick to the giant detailed demographic studies, or the lab-based work with good mechanistic support. As far as I can tell, there are actually some real changes in cancer, asthma, autism, with industrialization, but its hard to work out causation and magnitude and there are plenty of examples of genetic clusters once thought environmental, or known exposures to toxins without expected impacts. It's complicated.
July 11, 2014, 10:37 ammarque2:
Asthma as well. As them a cases went way up when a government program to alert people and doctors to the condition was started. Mostly people who had it and were formerly not diagnosed and those with marginal cases were added. We also has an expansion of the definition - cold air asthma, activity related asthma, etc.
Of course when the cases rose because the awareness campaign worked - Eco nuts claimed it was due to more pollution and diesel particulates even though those were being cleaned up. Now CO2 is blamed as well
July 11, 2014, 12:22 pm