I am currently doing multiple runs with different starting parameters to check identifiability. When using the combined error model, I got positive and negative b.1 values with very similar absolute values. I think it is preferable to restrict error model parameters to be positive. This PR implements this for the combined error model. I assume this is not a problem for the other error models, but did not check.
I am currently doing multiple runs with different starting parameters to check identifiability. When using the combined error model, I got positive and negative b.1 values with very similar absolute values. I think it is preferable to restrict error model parameters to be positive. This PR implements this for the combined error model. I assume this is not a problem for the other error models, but did not check.