For those intrigued by the Oregon Medicaid study!

TUESDAY, MAY 7, 2013

As all progressives should be: Yesterday, in our day of travel, we missed Kevin Drum’s post from late Sunday night about the Oregon Medicaid study.

If you’re trying to figure out that study, you should take a look at that post.

We’ll note that Drum was annoyed, as we were, by a claim which got its start in Annie Lowrey’s hapless news report in the New York Times. Late in his post, he says this:
DRUM (5/5/13): You can't honestly say that the study shows that Medicaid "seemed to have little or no impact on common medical conditions like hypertension and diabetes." That just isn't what it showed.
Drum was quoting Ross Douthat’s column in the Sunday Times. That said, Douthat was advancing a theme which got its start in the Times on Thursday morning, in Lowrey’s bungled report.

For our earlier post about Lowrey's report, click here.

In his post, Drum is examining a question we will try to examine tomorrow. He’s asking this: What should Lowrey and Douthat and others have said about this study’s findings? We don’t think we agree with every single word Drum says. But he’s clearly asking the right questions and his post is quite informative.

As best we can tell, this whole thing turns on the way reporters deal with a rather basic concept—“statistical significance.” As best we can tell, Lowrey dealt with the concept quite poorly, as we’ll see tomorrow. So did many others, including Douthat and the reporter at Slate to whom he links in his column.

In such ways, incompetent upper-end reporters help spread prime conservative bullroar. In this case, Douthat was spreading an idea which Lowrey also seemed to assert—the idea that Medicaid “seemed to have little or no impact on common medical conditions like hypertension and diabetes.”

To a certain type of conservative ear, that claim sounds like solid gold. More on this gong-show tomorrow—but if liberals and progressives care about topics like this, they should be upset when the New York Times hires reporters, editors and pundits who can’t do better than this.

Note to progressives: This fight will not be waged by the “career liberal” class. They’re too busy worrying about their own standing, present or future, with the glorious Times.

4 comments:

  1. Drum's point about a 20% improvement in GH level not being statistically significant is correct, according to my biostatistician wife.

    Another possible problem with the study comes from Drum's observation that:

    The study started out with about 6,000 people who were offered Medicaid. Of that, 1,500 actually signed up.

    The fact that only 25% of the people who were offered Medicaid chose to enter that program means that the study didn't look at a random sample. It's possible that the 1500 people who chose Medicaid may have been less healthy than the 4500 who turned it down. If so, then those on Medicaid might have been on average less healthy than the control group. In that case, the Medicaid group would have done well just to reach the health level of the control group after 2 years.

    I don't know that this actually is the case, but I don't see how it can be ruled out from what we know.

    ReplyDelete
    Replies
    1. The people running this study are not simple minded.

      They are looking at a) the entire population of people who registered for the lottery, b) the population selected in the lottery whether or not they eventually enrolled, as well as c) the population that enrolled.

      So they can control for this kind of self-selection bias.

      Delete
  2. No additional time expended on the gym, no back discomfort
    simply because of to numerous crunches or other stomach exercise routines and no far
    much more sweaty workouts merely to guarantee that your
    abs appear superb.

    my blog ... next Page

    ReplyDelete
  3. Although technically a subset of innumeracy, instatisticacy should be a recognized condition. "Statistical illiteracy" is too long. Instatisticacy should also be a recognized word.

    For the suggestible reader who would like to be less gullible: How to Lie with Statisics

    ReplyDelete