Parameter bounds and confidence intervals in LsqFit.jl?

[PeX]

HI all,

I’m trying to use LsqFit.jl to fit a model in the form of:

model(t, p) = @. p[1] * t^3 + p[2] * t + p[5]*sin(2*π*p[3]*t + p[4])

However, I need p[4] to be strictly between 0 and pi. Is it possible to do it?
In addition, is there a wat to calculate or plot the 95% confidence intervals of the fit?

If not, is there another recommended package that can deal with this effectively?

Thank you!

[abraemer]

@PeX

However, I need p[4] to be strictly between 0 and pi. Is it possible to do it?

I don’t think LsqFit.jl supports bounded optimization. However in this case it would not be too hard to transform the parameters found by the optimization to the desired interval.

@PeX

In addition, is there a wat to calculate or plot the 95% confidence intervals of the fit?

The Readme lists

confidence_interval = confint(fit; level=0.05, atol, rtol):

• fit: result of curve_fit (a LsqFitResult type)
• level: confidence level
• atol: absolute tolerance for negativity check
• rtol: relative tolerance for negativity check
  This returns confidence interval of each parameter at level significance level.

@PeX

If not, is there another recommended package that can deal with this effectively?

I think there are quite a few packages for optimization but I am not familiar with the ecosystem. Maybe someone else will make a recommendation.