Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- 1)a)
- You have to calculate the heat from an electrical resistor (constant in time).
- You have two technicians -
- technician A, who can measure voltage to a precision of 10mV
- technician B, who can measure amperage (?) to a precision of 0.07*10^(-4) A.
- technician A performed 10,000 tests and got V_{average} = 250[mV] with a stddev of 30[mV].
- technician A performed 10,000 tests and got I_{average} = 0.2*10^(-4)[A] with a stddev of 0.04*10^(-4)[A].
- You only have a budget for one of the following options:
- 1. buy a new device for technician A, 10 times more accurate, and perform one more test using the new device.
- 2. fund 10,000 more tests by technician A.
- 3. buy a new device for technician B, 10 times more accurate, and perform one more test using the new device.
- 4. fund 10,000 more tests by technician B.
- technician B will not reduce its error by performing more tests, because we take as the error max{stddev, instrument error}, and his precision is already higher than stddev.
- that isn't the case for technician A. we are left with options 1,2,3.
- if we perform one more test with a lower instrument error by technician A, the standard deviation will still be higher than the instrument error, and since we take the error to be max{stddev, instrument error}, this will not lower our error. we are left with options 2,3.
- if we perform one more test with an instrument error of 0.007*10^(-4)A, then the stddev will still be almost the same: 0.04*10^(-4)A. since we take the error to be max{stddev, instrument error}, reducing the instrument error will just make us take the stddev, of 0.04*10^(-4)A. it is multiplying the original error by ~0.571.
- we will get a relative error of roughly 20%.
- we know that error as a function of number of experiments N is error=cons * 1/sqrt(N), so performing another 10,000 experiments for each one, or twice as many, will multiply the stddev by 1/sqrt(2) ~== 0.707.
- we will get a relative error of roughly 8.5%
- therefore, we should fund 10,000 more tests by technician A to get a lower error.
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement