Question about the parameters of the fit

[Boris Deshev]

Hi, 

I am fitting the built-in SkewedGaussianModel to some data. The fit seems to be working well but when I print the parameters of the fitted model the values seems to be off. Looking at the attached plot I expect the amplitude to be around 30, the center around 6, sigma around 2, I guess. When I print the parameters I get

{'amplitude': 152.51586530547021, 'center': 4.95450125813345, 'sigma': 3.5260258280394083, 'gamma': 3.074519008172669}

What am I missing?

The code used to make this plot:

#-----

import numpy as np
import matplotlib.pyplot as plt
from lmfit.models import SkewedGaussianModel
np.random.seed(42)

bins = np.linspace(0, 15, 101)

# Generate something like a skewed Gaussian distribution
mult_x = np.linspace(1, 2, 1000)
data_y = (np.random.randn(1000) + 5) * mult_x

# Make the histogram
vals, bins = np.histogram(data_y, bins=bins)
x = (bins[1:]+bins[:-1])/2

# Fit a model
model = SkewedGaussianModel()
params = model.make_params(amplitude=np.max(vals),\
                           center=x[np.argmax(vals)],\
                           sigma=1,\
                           gamma=2)
Dgau = model.fit(vals, params=params, x=x)

# Get the fitted values
fitted_values = Dgau.eval(x=x)

# Plot
plt.plot(x, vals)
plt.plot(x,fitted_values)
plt.axvline(x=Dgau.values['center'], color='black', lw=0.5)
plt.show()
print(Dgau.values)

#-----

Cheers,

Boris

[Matt Newville]

On Sat, Dec 30, 2023 at 9:20 AM Boris Deshev <des...@gmail.com> wrote:

Hi, 

I am fitting the built-in SkewedGaussianModel to some data. The fit seems to be working well but when I print the parameters of the fitted model the values seems to be off. Looking at the attached plot I expect the amplitude to be around 30, the center around 6, sigma around 2, I guess. When I print the parameters I get

{'amplitude': 152.51586530547021, 'center': 4.95450125813345, 'sigma': 3.5260258280394083, 'gamma': 3.074519008172669}

What am I missing?

The definition used for SkewedGaussian is at https://lmfit.github.io/lmfit-py/builtin_models.html#skewedgaussianmodel

Amplitude, center, and sigma are those of the unskewed (unit-normalized) Gaussian.  That amplitude is not the maximum value of that Gaussian, it would be the area.    Similarly, the center would be the center of the Gaussian.  But that gets skewed by multiplying by an error function, centered at center, and with gamma and sigma controlling how sharp that rise is.

--Matt 

[Boris Deshev]

Got it. Thanks!

--
You received this message because you are subscribed to the Google Groups "lmfit-py" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lmfit-py+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lmfit-py/CA%2B7ESbrfU74Qsr2_3eKgym%2B5qSubW6%3DLrNnryKn%3DRVu6CJ7uGw%40mail.gmail.com.