_th1.pyzdoc

Go to the documentation of this file.

\pythondoc TH1
 
## Fitting histograms in Python
 
One-dimensional histograms can be fit in [Python](https://root.cern/manual/python) with a similar syntax as in C++.
To fit a 1D histogram to one of the ROOT standard functions (e.g. a Gaussian):
 
\code{.py}
# Create and initialize a test histogram to fit
myTH1D = ROOT.TH1D("th1d", "Histogram for fitting", 200, 0, 10)
myTH1D.FillRandom("gaus", 1000)
 
# Fit to a ROOT pre-defined Gaussian function "gaus"
myTH1D.Fit("gaus")
\endcode
 
The list of standard functions in ROOT can be accessed with the TROOT::GetListOfFunctions.
In Python, the standard functions for TF1 can be printed as follows:
 
\code{.py}
ROOT.TF1.InitStandardFunctions()
 
# Print a list of available functions and their definitions
ROOT.gROOT.GetListOfFunctions().Print()
\endcode
 
## Accessing results of the fit in Python
 
To access the results of the fit, run the TH1::Fit method with the "s" option (please see the TH1::Fit(TF1*, Option_t*, Option_t*, Double_t, Double_t)
documentation for a list of possible options).
This will return a TFitResult which can be examined with the corresponding TFitResult methods, with the same names in Python as in C++.
 
For example:
 
\code{.py}
# Re-using the TH1D defined in the earlier example code
myResult = myTH1D.Fit("gaus", "s")
 
# Get the fitted parameters as a vector
myResult.Parameters()
 
# Get the error of the first parameter
myResult.ParError(0) 
\endcode
 
 
## Fitting to user-defined functions in Python
1D histograms can also be fit to any user-defined function expressed as a TF1 (see the TF1 documentation for examples on how to do this).
 
For example, a TF1 can be defined and initialized with its ROOT constructor:
 
\code{.py}
# Define the function, e.g. a polynomial with two parameters: y(x) = a * x^b
myTF1 = ROOT.TF1("myFunction", "[0] * pow(x, [1])", 0, 10)
 
# Set parameters
myTF1.SetParameters(10.0, 4.0)
 
# Initialize a test histogram to fit, and fit it
myTH1D = ROOT.TH1D("th1d", "My histogram to fit", 200, 0, 10)
myTH1D.FillRandom("myFunction", 1000)
myTH1D.Fit("myFunction")
\endcode
 
A TF1 can also be defined using a Python function, for example:
 
\code{.py}
def myGaussian(x, pars):
    '''
    Defines a Gaussian function
    '''
    return pars[0]*np.exp(-0.5* pow(x[0] - pars[1], 2)) 
 
# Initialize from the Python function with the range -5 to +5, with two parameters to fit, and a one-dimensional input x
myTF1 = ROOT.TF1("myFunction", myGaussian, -5, 5, npar=2, ndim=1) 
 
# Create a 1D histogram and initialize it with the built-in ROOT Gaussian "gaus"
myTH1D = ROOT.TH1D("th1d", "Test", 200, -5, 5)
myTH1D.FillRandom("gaus", 1000)
 
# Fit the 1D histogram to our custom Python function
myTH1D.Fit("myFunction")
\endcode
 
## Pythonizations
The TH1 class has several additions for its use from Python, which are also available in its subclasses (e.g., TH1F, TH1D).
 
### In-Place Multiplication
 
TH1 instances support in-place multiplication with a scalar value using the `*=` operator:
 
\code{.py}
import ROOT
 
h = ROOT.TH1D("h", "h", 100, -10, 10)
h.FillRandom("gaus", 1000)
 
# Multiply histogram contents by 2
h *= 2
\endcode
 
This operation is equivalent to calling `h.Scale(2)`.
 
### Filling with NumPy Arrays
 
The Fill method has been pythonized to accept NumPy arrays as input. This allows for efficient filling of histograms with large datasets:
 
\code{.py}
import ROOT
import numpy as np
 
# Create a histogram
h = ROOT.TH1D("h", "h", 100, -10, 10)
 
# Create sample data
data = np.random.normal(0, 2, 10000)
 
# Fill histogram with data
h.Fill(data)
 
# Fill with weights
weights = np.ones_like(data) * 0.5
h.Fill(data, weights)
\endcode
 
The Fill method accepts the following arguments when used with NumPy arrays:
- First fill
- Second argument (optional): NumPy array containing the weights for each entry
 
<em>Please note</em> that when providing weights, the length of the weights array must match the length of the data array. If weights are not provided, all entries will have a weight of 1. A ValueError will be raised if the lengths don't match:
 
\code{.py}
# This will raise ValueError
data = np.array([1.0, 2.0, 3.0])
weights = np.array([0.5, 1.0])  # Wrong length!
h.Fill(data, weights)  # Raises ValueError: "Length mismatch: data length (3) != weights length (2)"
\endcode
 
The original Fill method functionality is preserved for non-NumPy arguments:
 
\code{.py}
# Traditional filling still works
h.Fill(1.0)  # Fill single value
h.Fill(1.0, 2.0)  # Fill single value with weight
\endcode
 
## Further Python fitting examples
Further examples can be found in the tutorials:
- [combinedFit.py](combinedFit_8py.html) performs a combined (simultaneous) fit of two 1D histograms with separate functions and some common parameters.
- [fit1.py](fit1_8py.html) reads a `TF1` and 1D histogram (created and saved in an earlier example [fillrandom.py](fillrandom__8py.html)), and fits the histogram.
- [fitConvolution.py](fitConvolution_8py.html) fits a 1D histogram to a convolution of two functions.
- [fitNormSum.py](fitNormSum_8py.html) fits a 1D histogram to the normalized sum of two functions (here, a background exponential and a crystal ball function).
- [multifit.py](multifit_8py.html) fits multiple functions to different ranges of a 1D histogram.
 
\endpythondoc