Skip Navigation
117 comments
  • Some bureaucrats in Mexico City tried this years ago.

    An important ring road had two lanes in each direction. To increase its capacity, they didn't actually widen the road; they just repainted the lane markings to turn two lanes into three, and claimed a 50% capacity increase!

    Everyone immediately screamed about being crammed together just centimetres apart, accidents increased and the city officials quickly u-turned; they repainted to have just 2 lanes in each direction again.

    But they then tried to claim that as that was a 33% decrease, and that because they had earlier increased it 50%, that meant they had achieved a net 17% increase in the road's capacity!

  • This is why in forecasting and time series analysis is used the log difference, a 10% increase or decrease on the log scale gives you the same value being added or removed.

    • Yes, that requires a reference value to be decided upon beforehand.

      • Not really, you do t=n and t=n+1, for n= 1, 2, 3 for a quick view on volatility.

        Then ypu look up for correlations between e[t=n | t= 0, t= 1...] for different Ns. For more I would need to check out my notes

117 comments