From appointment of our goverment the economy is growing 2x faster!!!
Really? Something is wrong with x-axis
If you do not use time scales of the same size on the x-axis (e.g. months), the shape of the graph can change significantly. As the manipulator in the left chart has stretched the 8 months of old government to a width twice as long as the 16 months of new government, economic growth seems steeper. What is the truth? The new government actually achieved an increase of 210 points, as did the old government, but for twice as long. And the new government probably wouldn’t brag about that.
This is how much our profits fallen. We must fire the people!!!
Really? Let’s try to use y-axis from 0 to 100%
For example, if someone wants to emphasize how much value has fallen over the years, they use the first graph. If not, he will use the second. And both are based on the same data and are basically both correct. However it is about context. Sometimes a 2% difference can be important. Sometimes not, and using the first graph, the good manipulator will certainly scares his audience easier.
What story does this picture want to tell me?
The next time someone shows you a graph, do not just pay attention to the commas and curves. Check:
Axes, numbers, scales and context.
THERE IS NO GLOBAL WARMING!
Let’s take a closer look
The situation may also change depending on the context. In the graph above, global temperatures appear to have increased only slightly in 140 years. In a context in which a change of one degree can disrupt the entire ecosystem of the planet and make part of the planet an uninhabitable area, the data need to be looked at more closely. That is why the chart below is much more informative.
How do the statistics lie?
Imagine two hospitals – X and Y.
Hospital X cured 900/1000 (90%) and Y cured 800/1000 (80%) of its patients.
Which hospital to go to? Hospital X cured more patients. I will go there!
Really?
Seriously ill: Hospital X has healed 30/100 – 30 %
Hospital Y healed 210/400 – 52,2 %
Slightly ill: Hospital X has healed 870/900 – 96,6 %
Hospital Y healed 590/600 – 98,3 %
This is strange! Hmmm…
Here is the explanation.
It is called the Simpson’s paradox.
The problem is that we have brought together two different groups: the seriously and the slightly ill. If you want to compare something like this, you must first think about whether you are not associating groups of people (subjects) with too different characteristics in the context of what you are researching (e.g. health of old and young people, prices of houses and flats, etc.).
| Seriously ill | Slightly ill | Overall | |
| Hospital X | 30/100 (30 %) | 870/900 (96,6 %) | 900/1000 (90 %) |
| Hospital Y | 210/400 (52,5 %) | 590/600 (98,3 %) | 800/1000 (80 %) |