4 ms·

Statistics is one field in particular where it's easy for non-experts to reach completely false conclusions.

nix23 6y ago
True with complex stuff, but we talk about covid-cases/day/region...that's really something everyone should interpret correctly.
- jjcon 6y ago
  COVID statistics can be very very difficult to interpret correctly, especially at the beginning when data was scarce. For instance, most people agree it is best to exclude totally random testing sites from COVID numbers because even though the false positive rate is very low, at the beginning so was the incidence of COVID which means most of the positive tests would be false positive. Including them or improperly dealing with that data would be considered base rate fallacy https://en.m.wikipedia.org/wiki/Base_rate_fallacy https://en.m.wikipedia.org/wiki/Base_rate_fallacy This is just one statistical phenomenon that has to be dealt with. It’s complicated and sometimes messy but it absolutely requires top minds
- SpicyLemonZest 6y ago
  Proper interpretation of even the simplest data is easy to get wrong. For example, I routinely see smart, well-informed people describe the number of Covid deaths reported in a day as the number of people who died of Covid that day.
- hedora 6y ago
  Yeah; some sibling comments here say Covid stats are difficult to interpret, but cite phenomena (non random samples, false positives, conflation of population and sample statistics, etc) that should be covered by any reasonable introductory statistics course. In the US, that’s first or second year undergrad territory. Overseas, it’s often covered in high school. It’s particularly depressing that science reporters regularly get these things wrong. I can’t imagine what (lack of) credentials are necessary to land such a job, but I often wonder if a basic understanding of statistics is a disqualifying credential.
  - hedora 6y ago
    (also, the flagging of nix23’s comment is way out of line.)
    - dang 6y ago
      If you mean https://news.ycombinator.com/item?id=25811860 https://news.ycombinator.com/item?id=25811860, the comment was correctly flagged because the flamebait in it broke the site guidelines badly.
  - jjcon 6y ago
    Mods this is obviously a nix23 alt account which was already flagged in multiple places by mods for spamming flame bait. Both accounts post in the same places regularly. But I’ll bite. The point of providing an example is to show how there can be unobvious effects that are specific to the medical field (and isn’t as simple as cases per day). It takes Doctors and statisticians working together to interpret this info correctly. And no not al the effects are covered in stats101 despite how the freshman may protest.
    - dang 6y ago
      I see zero evidence of that, and it's a serious violation of the site guidelines to post like that. Please see https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html. As it says there, if you're worried about abuse, email hn@ycombinator.com and we'll look at the data. One of the most important lessons I've learned from poring over this data is that 99% of the things that internet users describe as "obvious" are not only obvious but false. A strange phenomenon, and one we all need to work on overcoming if we're to have real conversation.

True with complex stuff, but we talk about covid-cases/day/region...that's really something everyone should interpret correctly.

jjcon 6y ago
COVID statistics can be very very difficult to interpret correctly, especially at the beginning when data was scarce. For instance, most people agree it is best to exclude totally random testing sites from COVID numbers because even though the false positive rate is very low, at the beginning so was the incidence of COVID which means most of the positive tests would be false positive. Including them or improperly dealing with that data would be considered base rate fallacy https://en.m.wikipedia.org/wiki/Base_rate_fallacy https://en.m.wikipedia.org/wiki/Base_rate_fallacy This is just one statistical phenomenon that has to be dealt with. It’s complicated and sometimes messy but it absolutely requires top minds
SpicyLemonZest 6y ago
Proper interpretation of even the simplest data is easy to get wrong. For example, I routinely see smart, well-informed people describe the number of Covid deaths reported in a day as the number of people who died of Covid that day.
hedora 6y ago
Yeah; some sibling comments here say Covid stats are difficult to interpret, but cite phenomena (non random samples, false positives, conflation of population and sample statistics, etc) that should be covered by any reasonable introductory statistics course. In the US, that’s first or second year undergrad territory. Overseas, it’s often covered in high school. It’s particularly depressing that science reporters regularly get these things wrong. I can’t imagine what (lack of) credentials are necessary to land such a job, but I often wonder if a basic understanding of statistics is a disqualifying credential.
- hedora 6y ago
  (also, the flagging of nix23’s comment is way out of line.)
  - dang 6y ago
    If you mean https://news.ycombinator.com/item?id=25811860 https://news.ycombinator.com/item?id=25811860, the comment was correctly flagged because the flamebait in it broke the site guidelines badly.
- jjcon 6y ago
  Mods this is obviously a nix23 alt account which was already flagged in multiple places by mods for spamming flame bait. Both accounts post in the same places regularly. But I’ll bite. The point of providing an example is to show how there can be unobvious effects that are specific to the medical field (and isn’t as simple as cases per day). It takes Doctors and statisticians working together to interpret this info correctly. And no not al the effects are covered in stats101 despite how the freshman may protest.
  - dang 6y ago
    I see zero evidence of that, and it's a serious violation of the site guidelines to post like that. Please see https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html. As it says there, if you're worried about abuse, email hn@ycombinator.com and we'll look at the data. One of the most important lessons I've learned from poring over this data is that 99% of the things that internet users describe as "obvious" are not only obvious but false. A strange phenomenon, and one we all need to work on overcoming if we're to have real conversation.