CalculateTransmission error propagation

[Gesner Passos]

While I was working on a ticket related to CalculateTransmission I've spotted an error on how we calculate the error propagation.

 // Take the log of each datapoint for fitting. Recalculate errors remembering that d(log(a))/da  = 1/a
      E[i] = std::abs(E[i]/Y[i]);
      Y[i] = std::log10(Y[i]);

In reality, d(log(a))/da = 1/(a*ln10) this means, that we should have this lines:

 // Take the log of each datapoint for fitting. Recalculate errors remembering that d(log(a))/da  = 1/(a*ln10)
      E[i] = 0.434*std::abs(E[i]/Y[i]);
      Y[i] = std::log10(Y[i]);

[Gesner Passos]

(In #8977) While doing this ticket I've spotted an issue with the error propagation #9011

[Nick Draper]

Bulk move of tickets out of triage (new) to assigned at the introduction of the triage state

[Gesner Passos]

The entire fit will change the way it calculates error. It will have impact on CalculateTransmission.

[Gesner Passos]

Script to show fitting in CalculateTransmission

[Gesner Passos]

Although the description is correct. It happens that when it is supposed to go back from the error it is using the correct transformation to go back.

If you use the check9011. It uses the transformations Log/Fit/Exponential and Log10/Fit/10^x to compare with the results from CalculateTransmission. And the answers are the same.

Hence, no reason to change.

[Stuart Campbell]

This ticket has been transferred to github issue ​9854