Comments on: Delta Surge Update – Demographics Focus 8/13/21

By: Tod Newman

Tod Newman — Fri, 08 Oct 2021 23:06:45 +0000

In reply to Winston Xia. Hi Winston, see my reply to Barclay on my thoughts of the challenge of understanding when to normalize and when not to...

By: Tod Newman

Tod Newman — Fri, 08 Oct 2021 23:06:01 +0000

In reply to Barclay Brown.

Thanks Barclay… here’s my thoughts on this one analytic… I’m constantly shifting between normalized data and non-normalized data based of off the data story that makes the most sense. Sometimes I show “deaths per 1000 people per day” as a rate, for instance, but I find that this is less relevant to people who aren’t attuned with statistics than “deaths per day”. I’m not happy with the latter because then I can’t compare, say, New York with Rhode Island, but it makes more sense broadly for communication.

I think that the analytic you’re referring to is one of these… I’m trying to show the ratio of deaths over age 65 to those under 65, this is an interesting metric that could benefit from normalization… but if I do, the ratio becomes too large to be understandable (the number of deaths for over 65 to the population of over 65 is HUGE) and conversely for everyone else it is really tiny. Therefore the ratio is too big to be reasonable in most minds.

This is a pretty interesting thing to think about… I definitely learned a lot about the divide of when to normalize or when not to from COVID.

By: Winston Xia

Winston Xia — Thu, 26 Aug 2021 12:02:04 +0000

In reply to Barclay Brown. Agree, makes more sense to compare using only the death rate “per thousand population” stats. Or “per ten thousand population “ Otherwise kind of misleading (using just the absolute death #, vaccination rate etc.)

By: Barclay Brown

Barclay Brown — Fri, 20 Aug 2021 11:19:29 +0000

Hi Tod, I love your articles and your approach to analysis. While you mention that the size of the various age groups is different and imply that this can have an effect, why not eliminate that effect entirely by using only “per thousand population” stats, e.g. deaths per thousand in the 65+ age group instead of absolute deaths. Without this, what’s the meaning of the ratio of deaths between groups? Say we have a group of 10 and 2 die for 20%. Then we have a group of 100 and 20 die–also 20%. But the ratio is 1:10 which seems pretty meaningless since the death rate is actually the same in both groups. I bet there are lots more deaths from accidents in Toyotas than from Teslas, but there are LOTS more Toyotas out there. Deaths per thousand cars of each make could tell a different story.