Garett Jones tells us that he got the idea from Joel Schneider who made the similar statement in 2007 that:
"I Am as Tall as the Rocky Mountains! (After controlling for barometric pressure)
Just because I can use regression to show that the houses in my neighborhood are just as tall as the Rocky Mountains after statistically controlling for barometric pressure at the summits of each house and mountain does not mean that I have proved that there is no true difference in the heights of houses and mountains. It also doesn’t prove that houses and mountains would be the same height if we were to equalize the barometric pressure differences. We know that such an analysis is stupid because we know that changing altitude causes air pressure to change and that changing air pressure has no effect on altitude."
Great post. I prefer the version, "Mt. Everest has the same elevation as London after controlling for air pressure."
The fallacy seems to be pretending that endogenous variables are exogenous. The reason why such researchers would control for unemployment rate etc. of migrants is that: "If we get them employed then they won't be so criminal." But of course both unemployment and criminality may follow largely from genetic variation.
I could not capture the real meaning of the "Mont Everest fallacy.". Just to make it sure, this fallacy occurs when we add predictors to regression models that shouldn't be added. Don't you have any statistical method to detect these predictors that should not be there beforehand, such as VIF ?
You need causal reasoning and models to apply regression models correctly. Causality cannot just be inferred from most data, so it is not possible just to apply VIF or any other statistical test.
What would cos do here? The point is the distance to equator obviously affects temperature. Absolute value of latitude is a kind of distance to equator metric.
It makes some sense from perspective of human tourist from %typicalhumandwelling% taking a tour to %placename% and determining beforehand which clothes they need. Fishes living in ocean are not allowed on Everest, sorry!
Garett Jones tells us that he got the idea from Joel Schneider who made the similar statement in 2007 that:
"I Am as Tall as the Rocky Mountains! (After controlling for barometric pressure)
Just because I can use regression to show that the houses in my neighborhood are just as tall as the Rocky Mountains after statistically controlling for barometric pressure at the summits of each house and mountain does not mean that I have proved that there is no true difference in the heights of houses and mountains. It also doesn’t prove that houses and mountains would be the same height if we were to equalize the barometric pressure differences. We know that such an analysis is stupid because we know that changing altitude causes air pressure to change and that changing air pressure has no effect on altitude."
http://www.iqscorner.com/2007/05/temp.html?m=1
Great post. I prefer the version, "Mt. Everest has the same elevation as London after controlling for air pressure."
The fallacy seems to be pretending that endogenous variables are exogenous. The reason why such researchers would control for unemployment rate etc. of migrants is that: "If we get them employed then they won't be so criminal." But of course both unemployment and criminality may follow largely from genetic variation.
"I will now use this model to encourage the resettlement of migrant populations on Mt. Everest."
There's no issue with the kitchen sink approach. There's a mighty big issue with interpreting the results and avoiding multicollinearity though.
To interpret Everest, we look at the results and see that Everest is indeed cold, mostly due to altitude.
I could not capture the real meaning of the "Mont Everest fallacy.". Just to make it sure, this fallacy occurs when we add predictors to regression models that shouldn't be added. Don't you have any statistical method to detect these predictors that should not be there beforehand, such as VIF ?
You need causal reasoning and models to apply regression models correctly. Causality cannot just be inferred from most data, so it is not possible just to apply VIF or any other statistical test.
is this you ? https://rationalwiki.org/wiki/Emil_O._W._Kirkegaard
nitpicky: why abs(lattitude) instead of cos or something?
What would cos do here? The point is the distance to equator obviously affects temperature. Absolute value of latitude is a kind of distance to equator metric.
Cosine makes more physical sense.
Anyways, thanks for great post!
Does Everest have a town? If not you're better off with a more random choice of locations?
Not really, but I wanted to illustrate with some known locations, not random spots on the world map. https://en.wikipedia.org/wiki/List_of_highest_towns_by_country
It makes some sense from perspective of human tourist from %typicalhumandwelling% taking a tour to %placename% and determining beforehand which clothes they need. Fishes living in ocean are not allowed on Everest, sorry!