Bias and imprecision calculation with $obsSD

[Trepon]

Hello,
In the Model bias and imprecision section of the Pmetrics manual (p. 53), the weighting calculation is described as follows:

For both calculations, the weighting is according to the calculated SD of the observation, using C0, C1, C2, and C3.

There is no mention of the error model (i.e. gamma/lambda) in the section.

When viewing the PMop object after a successful run, the values in the $obsSD column, which are used for the calculation of bias and imprecision, are described in the manual as (p.20):

Calculated standard deviation (error) of the observation based on the assay error polynomial

However, these values does not match the SD you can manually calculate with SD = C0 + C1*[obs] + C2*[obs]² + C3*[obs]³.
Instead, what is calculated is:

• For gamma error model, obsSD = SD*gamma (using the gamma value of the last cycle). This formula match the error formula for gamma error models described p.14:

error = SD * gamma

• For lambda error model, obsSD = SD + lambda (using the lambda value of the last cycle). However, this does not match the error formula for lambda:

error = (SD² + lambda²)⁰.⁵

So my questions are:

1. Should model bias and imprecision calculation only use SD or the whole error term (SD and gamma/lambda)?
2. If the error term has to be used, then there seems to be a mistake in the calculation for lambda error models?

Here is an example including a dataset, 2 models (gamma and lambda), the PMop objects generated and the final gamma/lambda values for each run:
example.zip

[Trepon]

Any update on this issue? I would like to know if I can rely on the displayed bias and imprecision values or if have to manually calculate them (using SD and gamma/lambda or only SD?).