I have a dataset containing the measured coordinates x, y plus an additional measured observable V.
Assuming that my model f=f(x) gives a prediction of the observable V, how can I implement my model such that the most constraining data points are the ones whose measure of y is compatible with a certain value (in the case of Gaussian errors)?
So far I’ve been using this, in the case my value is zero:
pm.Normal(name='Y', mu=0., sigma=obs_y_err, observed=obs_y)
Is this the correct approach?