Which statistical tools are best suited for this problem:
Team A goes and measures 3 magnitudes in 200 locations (let's imagine the volume of a room, the insulation coefficient and the average temperature outside the room). For each room they calculate the energy this room will consume to keep the desired temperature for a week.
So after the measurement campaign they have measured these 3 magnitudes in 200 different rooms (each room is only measured once). So the data looks like:
(Room_volume, r_coefficient, T, estimated_energy)
Team A is assumed to be very good at getting their results. Now team B measures the same 200 rooms, and calculates all the estimated energies for each room.
How should we measure the performance of team B vs team A (where A is assumed to be accurate)?
Forgetting the physics, as it's just an illustrative example, how do you compare the accuracy of a measurement campaign measuring 3 values from which a 4th one can be calculated deterministically? Note that none of these values are distributed in any special way among the 200 samples (they don't look gaussian)
I started looking at the distribution of errors, so:
{Estimated_energy_Ai - Estimated_energy_Bi} with i = 1..200
And seeing how many results are within a standard deviation. What other tools are appropriate?
