Is this a censored-data problem or not?

Thanks, Ricardo.

I have done a small analysis with synthetic data in the meantime, which also shows a bias when discarding measurements as described.

Do you have any pointers for how to model the truncation ‘in the middle’ for e, i.e. how to model measurements for e = 0 and e \ge 4 but not for e = 1, 2, 3?