Post #3814657

2026-07-14 05:36 UTC

@freshstart@hachyderm.io @ColinTheMathmo@mathstodon.xyz The graph is not a "standard deviation" (SD) - that statement doesn't even make sense. An SD is a measure taken from a normal (gaussian) distribution. As the title states, the vertical *scale* is measured in SDs from 1991 - 2020 mean. I assume they modelled temperature variation from 1991 - 2020 as a normal distribution and this is then a comparison to that period. The red line showing the current variation is way outside the temperature variation from that period.

Replies (1)

@bradr@infosec.exchange 2026-07-14 17:51

@level98@mastodon.social @freshstart@hachyderm.io @ColinTheMathmo@mathstodon.xyz The graph shows a standard z-score. The y-axis is not the magnitude of any standard deviation, rather it is the deviation (positive or negative) of that day's temperature from the base period's average temperature for that day (in units of the s.d. of base period temperatures for that day). This (among other things) corrects for seasonality, since the s.d. varies throughout the year. (math example: if during the base period from 1991 to 2020, day 150 had a mean of 10°C and a s.d. of 2°C, and day 150 of 2026 was measured to be 14°C, then the graph would show a y value for the 2026 line on day 150 of (14-10)/2 = +2 s.d.)

Open ##3814656