search

CorrelAid / paris-bikes

Where to build new bicycle parking spots in Paris? Supporting data-driven decision making with open data
MIT License
8 stars 1 forks source link

Normalize metrics by area #63

Closed operte closed 1 year ago

operte commented 1 year ago

Benoit suggested we normalize metrics by area, since the area of each IRIS varies considerably.

What do you think?

katoss commented 1 year ago

mhh I don't have a strong opinion. Maybe it helps to imagine a scenario. So I think of.. IRIS A (very big), and IRIS B (very small) have the same school capacity, number of shops, etc = same score. What does that mean for their demand in bike parking? Wouldn't it stay the same, independent of the area?

mpadge commented 1 year ago

There's no straightfoward answer to this question. Sometimes it is important to normalise by area, particularly for active transport because area scales with distance, which has a definite effect. Sometimes it is also equally important to scale for local population densities. And sometimes it is important to scale for both. My instinctive guess would be that the notional attempt to scale IRIS polygons to equal populations should generally mean that post-hoc scaling shouldn't be done.

operte commented 1 year ago

Ok, I think we should drop it and let the city work on it as they see fit :) Thanks everyone!