Skip to content

Scandal! Mister P appears in British tabloid.

Tim Morris points us to this news article:

And here’s the kicker:

Mister P.

Not quite as cool as the time I was mentioned in Private Eye, but it’s still pretty satisfying.

My next goal: Getting a mention in Sports Illustrated. (More on this soon.)

In all seriousness, it’s so cool when methods that my collaborators and I have developed are just out there, for anyone to use. I only wish Tom Little were around to see it happening.

P.S. Some commenters are skeptical, though:

I agree that polls can be wrong. The issue is not so much the size of the sample but rather that the sample can be unrepresentative. But I do think that polls provide some information; it’s better than just guessing.

P.P.S. Unrelatedly, Morris wrote, with Ian White and Michael Crowther, this article on using simulation studies to evaluate statistical methods.

Fake-data simulation. Yeah.


  1. Simon says:

    Also appears as part of Remain United’s methodology here (

    “However, a new approach has been gaining attention in the polling industry. It uses advanced statistical techniques of regression to provide a better way of handling unrepresentative samples. Already well-known in the technology industry, where it is called “machine learning”, it is a key part of algorithms used for internet search and language translation. Market researchers have given it the clunky name of multi-level regression and post-stratification, or MRP for short, but it is essentially the same technique.”


    • I had to go read the original to see if that quote was accurate. Sadly, it is.

      I guess they didn’t get the memo from Google rebranding machine learning as artificial intelligence, with a wide enough marketing scope to include linear regression. It’s amusing to think of Gauss as a machine learning researcher in the pre-machine days. I wonder if his NeurIPS papers would be accepted or if they’d reject them for being written with quill and ink.

      • Joe Hoover says:

        They might have missed Google’s memo, but not the zeitgeist:

        ‘The algorithm “learns” how people with particular characteristics are more likely
        to think. […] Once the learning stage is over the algorithm then gets to work on anonymous and aggregated census
        data, which shows how many people with each characteristic live in any area. It can estimate the
        political views of those people using the information it discovered earlier. Since the census is
        representative, this gives a final estimate of public opinion which should be more accurate.’

        Estimation is so much easier with algorithms that know when to learn and when to just get to work.

    • Andrew says:

      Hey, I guess that means I’m a market researcher!

  2. Bob says:

    Gee I wonder if that’s the result they hoped to get?

  3. BenK says:

    Polls do provide some information! And that’s what makes Arfur’s critique relevant – one question is ‘from how many polls was this result drawn?’
    If the story has an unreliable narrator, data is always anecdotal.

    • Bob says:

      Given that a single issue, pro Brexit party was set up a month ago, and is polling in the high thirties, I’d hold off a couple of weeks until the out of sample data comes in before getting too excited.

      In fact, I prefer the Taleb/ Harry Crane approach to this, anyone who doesn’t back their predictions with hard cash isn’t credible.

  4. John Williams says:

    Here is a question. The website “Data for Progress” talks about using MRP. Are Any of the people in that group your students?

  5. Tim Morris says:

    Thanks for the PPS, Andrew. That’s probably given our paper more exposure than all my other efforts!

  6. Robert Grant says:

    They (focaldata) might have even used Stan. Someone involved in Stan might have done some consultancy with them about Stan and MRP a couple of years back, though NDA prevents him/her from commenting further. Nice people, interesting times for pollsters.

  7. Bob says:

    ‘Voters back a second referendum in every area of the UK’. Hmmm…. Did the election results bear this out or not?

    1. Did these researchers back their view with real money?

    2. Partisan ‘research’ is partisan and untrustworthy, no matter how much multi level post stratification blah blah you use.

Leave a Reply to BenK