• Posts
  • RSS
  • ◂◂RSS
  • Contact

  • Publication Bias

    April 17th, 2012
    science  [html]
    If I correctly predicted the overall direction of the stock market (or a specific stock) five days in a row, that would be pretty good evidence that I've got my finger on the market's pulse. Getting that right with random guessing would only happen about once every 32 (2 to the 5th) times, so isn't this p=0.031? Wouldn't you let me manage your money, winking at my "past performance is no guarantee of future results" disclaimer?

    Let's say I start out with 32 people. I pick a stock. I tell half of them it will go up, the other half it will go down. Tomorrow 16 people will have seen me be right, 16 will have seen me be wrong. I give up on the second half and pick a second stock for day two. Of the 16 who have never seen me be wrong I tell 8 it will go up and 8 it will go down. I repeat this until day six when I have one person who has seen me be right the past five days.

    This is an illegal con, but something similar happens in studies. I claim "Foobaritol lowers blood pressure in tall people", but maybe I ran studies on several different drugs, only publishing the one with the strongest results. How do you know I didn't run the numbers on several ways different ways of dividing up my study participants (age, gender, location, race, height, ...) and am able to make my claim about tall people only because I sneakily neglected to make a Bonferroni correction?

    What if we required preregistration as a condition for IRB approval?

    Comment via: google plus, facebook

    Recent posts on blogs I like:

    What should we do about network-effect monopolies?

    Many large companies today are software monopolies that give their product away for free to get monopoly status, then do horrible things. Can we do anything about this?

    via benkuhn.net July 5, 2020

    More on the Deutschlandtakt

    The Deutschlandtakt plans are out now. They cover investment through 2040, but even beforehand, there’s a plan for something like a national integrated timetable by 2030, with trains connecting the major cities every 30 minutes rather than hourly. But the…

    via Pedestrian Observations July 1, 2020

    How do cars fare in crash tests they're not specifically optimized for?

    Any time you have a benchmark that gets taken seriously, some people will start gaming the benchmark. Some famous examples in computing are the CPU benchmark specfp and video game benchmarks. With specfp, Sun managed to increase its score on 179.art (a su…

    via Posts on Dan Luu June 30, 2020

    more     (via openring)


  • Posts
  • RSS
  • ◂◂RSS
  • Contact