Irreproducible fit failure with fitStatus=3

[yus]

Hi,

I’m fitting a gaussian in a rather large framework. It’s a pretty simple fit: we don’t provide the initial-parameter values letting ROOT to set them automatically for this predefined gaus function, and we don’t even provide the min/max values along the X that would limit the fitting range. The problem is that sometimes the fit fails with fitStatus=3. That is, the fitStatus defined as

const int fitStatus = h->Fit(...);

equals 3. While working on an MWE, I fail to reproduce the fit failure, it always succeeds. I’m using almost the same code as in the large framework, the same input h histogram, the same fit options, and the same ROOT version. My question: what does fitStatus=3 mean? I know it may depend on the minimizer: since I’m using ROOT 6.32.08, I believe my minimizer is Minuit2.

Thanks!

---

ROOT Version: 6.32.08
Platform: lxplus9
Compiler: Not Provided

---

[ferhue]

Might be: Edm is above max

The best would be to re-run it in your original framework, but with the verbosity option turned on (“V”) or with “Q” removed, then you would get more informative error messages.

[StephanH]

Hello,

if somebody lands here via the search, the fit status is explained in the documentation of TH1::Fit.

Curiously, status 3 is not explained there, therefore @ferhue suggestion of rerunning with higher verbosity is probably the best course of action.