Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: [Phys-L] Analysis of Half-Life measurement using PyStan

On 10/12/21 2:03 PM, Paul Nord wrote:
One can measure the background independently


and then subtract
that from the data before fitting.

Don't do that.

We are comparing the model to the data.
It is perfectly OK to add the measured background as a constant
(not fitted) parameter on the *model* side of the equation.

Subtracting it from the data side of the equation is not
equivalent. It screws up the probability contours, and
therefore screws up the fit.

Doing it right is zero additional work compared to doing it wrong.

I suspect you are being betrayed by your intuition based on
experience with Gaussians. You can shift a Gaussian left or
right and it's still a Gaussian. Not so with Poisson processes.

If this explanation is not satisfactory please let me know
which part needs to be clarified.