blog/reference-point: Add

Jakub Valenta 2 years ago
parent 3f5c614267
commit 750157fe13
  1. 3
  2. 3
  3. 3
  4. 77

@ -0,0 +1,3 @@
oid sha256:f5c6fa79c086629639c1e85e1938912ea8f1f3153e60796751583ca79018ac53
size 260015

@ -0,0 +1,3 @@
oid sha256:57d3f1edfe5a6142250f354dfcd0907cf910ff603267ecf103ac0bbcf88d06a5
size 152610

@ -0,0 +1,3 @@
oid sha256:4c87ed50155968d74d310b01d86f0f3d9bef0bd3d63bfa417c66906b390cf68c
size 261094

@ -0,0 +1,77 @@
title = "Give me a reference point and I will tell you how bad your incidence is"
template = "post.html"
meta_image = "blog/reference-point/covid-map-germany-yellow-red--1200x630.png"
_a commentary on how one dispute about colors showed deeper problems with Corona
<!-- more -->
There was a little argument between two major German news media some days
ago. It's related to Corona and data visualization and I have a clever answer,
so hear me out or scroll fast. Still here? All right, let's do this:
The story is that Bild
Tagesschau of showing manipulative maps of Covid incidence in German
regions. Bild noticed that on one day Tagesschau published a yellow map, but
months later they published a map with almost the same numbers but this time in
{{ image(path='blog/reference-point/covid-map-germany-yellow-red.svg', width=600, alt='Illustration of Covid incidence maps of Germany in yellow and red color') }}
Bild concluded that by choosing a scary color, Tagesschau tried to make the
pandemic look worse (because politics). Tagesschau
that one map version was actually for TV and the other for web and it just got
mixed up. The story could end here. But there's a deeper problem...
Think: How can we in general tell if a number is small or big or scary or ok?
It says nothing when a number is presented like "look, here's 1.5". We need a
reference point such as "here's 1.5 and by the way it was 1000 the whole last
year". Now we can tell something: 1.5 looks small.
So if a Corona map (like the Tagesschau's) wants to tell something about the
incidence numbers... where is the reference point? What are we supposed to
compare the numbers with to understand them? This might seem like a rhetorical
question suggesting there isn't such point, but no, there is:
The reference point for a number in one region are the numbers in other
regions. We can compare: Berlin — larger, Flensburg — smaller. Notice that it
doesn't even matter if the map is yellow, red or say blue, as long as the shades
differ. So reference problem solved? Not quite.
See, regional differences are not the only thing readers look for in the
map. They also ask: "How is Germany doing as a whole country?" But in the
Tagesschau map, a reference point to allow assessing Germany as a whole is
missing. This is the core of the problem.
{{ image(path='blog/reference-point/covid-map-germany-reference-points.svg', width=600, alt='Illustration of reference points in Covid incidencemaps in Germany') }}
The "how Germany is doing as a whole" information is represented by the average
color of the map. But what can we compare this color to? 1) Other countries?
Nope, there's just an ocean of grey. 2) The map from yesterday? Yeah, sure, do
you remember it?
The only thing we can compare the average color of the map to is a cultural norm
of which colors are considered good and which bad. But that's terribly vague and
destined to cause disagreements; exactly like in the Bild x Tagesschau case.
Summary: Tagesschau (and others) show maps that are only good for comparing
Covid incidence among regions and that fail to tell how the country as a whole
is doing, because there is no reference point to which we could compare the
average color; just a vague feeling of which colors are good and which bad.
Ideas for improvement? Glad you asked! 1) Show the neighbouring countries. 2)
Show a change over time. 3) Cite a source that assesses the numbers (e.g. the
health ministry, which already says which incidence numbers are good and which
bad — that's the traffic light system).
Funnily enough, it's Bild who published a
that does this better as it seems to use the Corona traffic light colors. Now
they just need to state explicitly that this is the source of the colors and the
reference point to which the regional incidence numbers should be compared to.