Phitter Docs

Appearance

Create Discrete Fit ​

To perform a discrete fit, the Phitter class must be instantiated. The class accepts several parameters, described below:

  • data (list[int | float] or numpy.ndarray):
    Dataset to which discrete probability distributions will be fitted.

  • fit_type (str, optional):
    Specifies the fitting approach. Use "discrete" when dealing with discrete data (default is "continuous").

  • confidence_level (float, optional):
    Confidence level employed in the goodness-of-fit tests for evaluating distribution adequacy. Default is 0.95.

  • minimum_sse (float, optional):
    Threshold of minimum sum of squared errors (SSE) for including distributions in the fitting results. Default is numpy.inf (no restriction).

  • subsample_size (int, optional):
    Size of a randomly drawn subset from the dataset for analysis. Default is None, meaning the entire dataset is used.

  • subsample_estimation_size (int, optional):
    Size of a random subset utilized specifically for parameter estimation of the distributions. Default is None.

  • distributions_to_fit (list[str] or "all", optional):
    List specifying discrete distributions to consider in the analysis. Refer to the Discrete Distributions Documentation for available identifiers. The default value "all" fits all available discrete distributions.

  • exclude_distributions (list[str] or "any", optional):
    Identifies distributions explicitly excluded from the fitting procedure. Default is "any", indicating no exclusions.

Basic Discrete Fit Implementation ​

A basic discrete distribution fit using default parameters is illustrated below:

python
import phitter

# Define the dataset
data: list[int | float] = [...]

# Instantiate and perform discrete fit
phi = phitter.Phitter(
    data=data,
    fit_type="discrete"
)
phi.fit(n_workers=4)

Extended Parameter Configuration ​

For a detailed discrete distribution fit specifying particular parameters, the following implementation can be used:

python
import phitter

# Define the dataset
data: list[int | float] = [...]

# Perform discrete fit with additional settings
phi = phitter.Phitter(
    data=data,
    fit_type="discrete",
    confidence_level=0.95,
    minimum_sse=1e-2,
    distributions_to_fit=["binomial", "geometric"]
)
phi.fit(n_workers=4)

Upon completing the fitting process, further evaluation of distribution fitting results can be carried out using methods documented in subsequent sections.