# What’s the Deal With Bayesian Statistics?

The Rev. Thomas Bayes, whose simple and fairly narrow formula eventually inspired a vast new field of statistics.

It is really, really hard to find stuff to write about other than the C19 pandemic. So instead, how about a complicated post about Bayesian statistics? The underlying study that prompted this post is, unfortunately, about coronavirus testing, but it’s really about Bayesian statistics. Honest.

Bad news first: most of you probably don’t know what Bayesian statistics is, and it would take way too long to explain properly. I’m not even sure I could do a fair job of it, actually, so just skip this whole post if it’s not your thing. OTOH, if you’re really bored, go ahead and read it and then use the internet machine to enlighten yourself about the whole subject.

The whole thing starts with that Santa Clara study of C19 antibody testing. The authors concluded that the infection rate of the entire population was between 2.5 percent and 4.2 percent, which includes even those who had no symptoms and never knew they were infected. Unfortunately, the antibody test had a false positive rate of 1.5 percent, which means that the true range of the infection rate was about 0 to 5 percent.

So were the authors justified in using their narrower range of 2.5 to 4.2 percent? The loud and clear voice of the statistics community was no. But today, Andrew Gelman, who was quite critical of the study, writes this:

It seems clear to me that the authors of that study had reasons for believing their claims, even before the data came in. They viewed their study as confirmation of their existing beliefs. They had good reasons, from their perspective.

….It’s a Bayesian thing. Part of Bayesian reasoning is to think like a Bayesian; another part is to assess other people’s conclusions as if they are Bayesians and use this to deduce their priors. I’m not saying that other researchers are Bayesian—indeed I’m not always so Bayesian myself—rather, I’m arguing that looking at inferences from this implicit Bayesian perspective can be helpful, in the same way that economists can look at people’s decisions and deduce their implicit utilities.

This has always been my big problem with Bayesian statistics and Gelman makes it very clear in this post. Bayesiansim requires you to start with “prior beliefs” as part of the methodology. That can be a “flat prior,” in which you assume nothing, but then you end up with little more than you’d get from ordinary frequentist statistics. On the other hand, if you assume a meaningful prior, it means that your results can be stretched into almost anything you want. You can simply explain, in whatever detail you want, that “previous studies have shown X,” and when you plug that in as your prior you get a new result of X ± Δ.

But as Gelman acknowledges, this makes “prior” just a nicer word for “bias.” That’s not a very good selling point. At the same time, without priors it’s not clear to me how useful Bayesian statistics is in the first place. For now, then, I remain confused.

### WE'LL BE BLUNT

It is astonishingly hard keeping a newsroom afloat these days, and we need to raise \$253,000 in online donations quickly, by October 7.

The short of it: Last year, we had to cut \$1 million from our budget so we could have any chance of breaking even by the time our fiscal year ended in June. And despite a huge rally from so many of you leading up to the deadline, we still came up a bit short on the whole. We can’t let that happen again. We have no wiggle room to begin with, and now we have a hole to dig out of.

Readers also told us to just give it to you straight when we need to ask for your support, and seeing how matter-of-factly explaining our inner workings, our challenges and finances, can bring more of you in has been a real silver lining. So our online membership lead, Brian, lays it all out for you in his personal, insider account (that literally puts his skin in the game!) of how urgent things are right now.

The upshot: Being able to rally \$253,000 in donations over these next few weeks is vitally important simply because it is the number that keeps us right on track, helping make sure we don't end up with a bigger gap than can be filled again, helping us avoid any significant (and knowable) cash-flow crunches for now. We used to be more nonchalant about coming up short this time of year, thinking we can make it by the time June rolls around. Not anymore.

Because the in-depth journalism on underreported beats and unique perspectives on the daily news you turn to Mother Jones for is only possible because readers fund us. Corporations and powerful people with deep pockets will never sustain the type of journalism we exist to do. The only investors who won’t let independent, investigative journalism down are the people who actually care about its future—you.

And we need readers to show up for us big time—again.

Getting just 10 percent of the people who care enough about our work to be reading this blurb to part with a few bucks would be utterly transformative for us, and that's very much what we need to keep charging hard in this financially uncertain, high-stakes year.

If you can right now, please support the journalism you get from Mother Jones with a donation at whatever amount works for you. And please do it now, before you move on to whatever you're about to do next and think maybe you'll get to it later, because every gift matters and we really need to see a strong response if we're going to raise the \$253,000 we need in less than three weeks.

### WE'LL BE BLUNT

It is astonishingly hard keeping a newsroom afloat these days, and we need to raise \$253,000 in online donations quickly, by October 7.

The short of it: Last year, we had to cut \$1 million from our budget so we could have any chance of breaking even by the time our fiscal year ended in June. And despite a huge rally from so many of you leading up to the deadline, we still came up a bit short on the whole. We can’t let that happen again. We have no wiggle room to begin with, and now we have a hole to dig out of.

Readers also told us to just give it to you straight when we need to ask for your support, and seeing how matter-of-factly explaining our inner workings, our challenges and finances, can bring more of you in has been a real silver lining. So our online membership lead, Brian, lays it all out for you in his personal, insider account (that literally puts his skin in the game!) of how urgent things are right now.

The upshot: Being able to rally \$253,000 in donations over these next few weeks is vitally important simply because it is the number that keeps us right on track, helping make sure we don't end up with a bigger gap than can be filled again, helping us avoid any significant (and knowable) cash-flow crunches for now. We used to be more nonchalant about coming up short this time of year, thinking we can make it by the time June rolls around. Not anymore.

Because the in-depth journalism on underreported beats and unique perspectives on the daily news you turn to Mother Jones for is only possible because readers fund us. Corporations and powerful people with deep pockets will never sustain the type of journalism we exist to do. The only investors who won’t let independent, investigative journalism down are the people who actually care about its future—you.

And we need readers to show up for us big time—again.

Getting just 10 percent of the people who care enough about our work to be reading this blurb to part with a few bucks would be utterly transformative for us, and that's very much what we need to keep charging hard in this financially uncertain, high-stakes year.

If you can right now, please support the journalism you get from Mother Jones with a donation at whatever amount works for you. And please do it now, before you move on to whatever you're about to do next and think maybe you'll get to it later, because every gift matters and we really need to see a strong response if we're going to raise the \$253,000 we need in less than three weeks.