Wrong plots when fitting a Weibull LN(dependent, independent) model

[ahaselsteiner]

I'm submitting a ...

• [x] bug report
• [ ] feature request
• [ ] code improvement request

Expected behavior

How should it work (with the bug fixed or the feature implemented)? The fits should get displayed correctly.

Screenshot

Actual behavior

How does it currently work (with the bug causing problems or without the feature)? The fits are plotted wrongly. The mu values are too low, precisely ln(correct value). The "fit of fits" plot is correct, however. The equation for mu is correct too.

Screenshot Wrong fit and correct "fit of fits":

Used model and correct equation:

Steps to reproduce the problem (how to see the actual behavior)

1. Upload the standard Vanem data set
2. Fit a Weibull, Log-normal(dependent, independent) model to it
3. Inspect the fit

[ahaselsteiner]

This is likely an error caused by the viroconcom package since the LognormalDistribution instance returns the same values for mu and scale. The mu value is correct, the scale value incorrect.

LognormalDistribution

[ahaselsteiner]

Interestingly, if the parameter is independent the mu value is correct.

This ist the output of LN LN(power, exp) model:

[ahaselsteiner]

I narrowed the problem further down.

When a mu is dependent the BasicFit objects are incorrect. They have scale value, which is too low. It is ln(scale) of the correct scale value.

[ahaselsteiner]

When the BasicFits objects are cretated, they are correct:

[ahaselsteiner]

The Fitting module seems to send correct BasicFits. However, they seem to get altered later:

[ahaselsteiner]

I think I got it:

[ahaselsteiner]

Yes, that was the problem.

Solved it:

[ahaselsteiner]

This is fixed. It was fixed on Master.