even cross the expected baseline, should make us refrain from making statements about untypical
mortality rate.
Large cruise ships are sometimes called ”floating nursing homes”. It might seem inappropriate
or offensive, but in many cases, it is what they really are. Pensioners often choose cruise ships over
residential elder-care facilities for better care quality and lower price.
This metaphor comes handy to have another common sense view at the numbers. Numerous
reports from pre-coronavirus crisis state that colds and flues are often overlooked despite being
one of the biggest threat in nursing homes, where it’s not uncommon that a single flu infection
outbreak is fatal for 5% of residents. [43]
The comparison to the flu actually turns out relevant. We can compare the Diamond Princess
incident to influenza outbreaks on cruise ships that happened in the past.
For example in September 2000 an outbreak of influenza-like illness was reported on a cruise ship
sailing between Sydney and Noumea with over 1,100 passengers and 400 crew on board. Laboratory
testing of passengers and crew indicated that both influenza A and B had been circulating on the
ship.
”Of the 1159 passengers on the cruise, 366 (32%) sought medical attention at the ship’s clinic
between days 1 and 13. Of these, 203 (55%), or 18% of all passengers, presented with respiratory
tract illness. Five patients had a primary diagnosis of pneumonia. Of the 203 passengers presenting
to the ship’s clinic with respiratory tract illness, 60 were identified as suspected cases on medical
record review by the ship’s doctor. 40 passengers hospitalized, two of whom died” [35]
The influenza incident actually gives a higher rate of hospitalized patients than on Diamond
Princess.
2.7
Mediator variable
It happens that experimental data proves causality between two variables but it does not give
insight to understand the relationship. Mediator variable is introduced to split the original rela-
tionship into the chain of cause-and-effect links. It allows clarify the nature of the relationships.
It might seem a bit abstract, so let’s proceed with an example.
The Aztecs believed that solar eclipse preceded the apocalypse, so they practiced ritual human
sacrifice to appease the gods and save the world. There is an evidence that eclipses led the Aztecs
to kill slaves and prisoners. If we could ”generate” more eclipses, we would have observed more
massive homicides. Is it fair to say that eclipse caused the massacres? Could a distant neutral
cosmic phenomenon cause manslaughter?
The relationship makes more sense if we introduce a mediator variable and realize that it was
the Aztecs’ fear and belief system that caused killings. Then we get a chain of two valid causal
relationships:
• Solar eclipse caused people’s fear
• People’s fear caused killings
In reasoning, sometimes we need to go deep to find the underlying cause. At times, on the
contrary, we must search for the more direct cause. Otherwise, if we trace back the cause-and-effect
relationships too far, we will find the cosmological Big Bang to be the cause of all events which is
not very helpful.
Identifying mediator variables is important in making rational decisions. Without understand-
ing that human fear is the more direct cause, the hypothetical progressive Aztecs had not been
able to make a right conclusion, that in order to prevent eclipse inspired homicide, they needed to
educate people and eradicate religion and superstition.
9