Walkabout wednesday

Why don't epidemiologists use model simplification?

My Health

This is a nice little video on why you should regard epidemiological studies with a grain of salt (or a rock ...). I wish someone would explain to me why epidemiologists don't use model simplification approaches when using multiple regression to control for confounding variables.

This is an interesting story about how statistical errors can propagate into bad policy recommendations. This is really the difference between 95% confidence limits on the mean response, and 95% prediction limits on the response. The upshot is that you might need alot more Vitamin D than previously thought. [ht: Rick Tyre]

Are your medications safe? Maybe, but there are big issues with how the FDA handles evidence of scientific misconduct in clinical trials. Basically, when they find evidence of misconduct that taints evidence in a trial, they sit on it or even actively seek to cover it up. As a result, drugs go on the market, or generics continue to be labeled as bio-equivalent, when it is not at all clear they should be. And we wonder why the public is starting to mistrust Science™.

You might have seen the news: dietary cholesterol isn't bad. This is a nice little blog post with some good thoughts on cholesterol, eggs, and food in general.

Your Education

When I first saw this interactive display of freshman beliefs over time I got very excited. Over the years the number of students applying to 6 or more colleges has soared; 10 times as many in 2014 compared to 1972. That's a lot of application fees. And yet, 55% of students are accepted by their first choice. That's a lot of unnecessary college applications out there.

And this! A illustrated taxonomy of all the ways you can be wrong. Ugh. I'm sure I'm guilty of a few of these every time I hit "publish".

Our Environment

More than 30% of freshwater invertebrates assessed by the IUCN are classified as "Data-deficient". Lucie Bland, now at the University of Melbourne, looked to see if she could predict the category of DD species from the limited data that is available. Then she put that into a decision theory framework to see if it was worth spending money to improve that information. This is both cool (because STATISTICS) and depressing. Depressing because she estimated that all DD mammal species could be completely surveyed for only $300 million, and using her algorithms the cost could be reduced by 69%. What is that depressing? Because a price tag of over $100 million is still way too big, although a drop in the bucket for the developed nations of the world.

And another from Environmental Decisions Group, the Huge Possum himself says, sometimes we should prefer restoration to protection.

And looks like there's some data to support the hypothesis that hypoxia at the top ends of reservoirs kills baby sturgeon and prevents natural recruitment. That's consistent with information on drift distances (the little devils almost for sure will drift into a reservoir), and the fact that slightly older hatchery reared fish survive. Now, what's going on in the lower Missouri River? [ht: Jason DeBoer]