size_matched_model.py

​

High-level description

The code defines a SizeMatchedModel class that represents a statistical model where parameters vary based on the size of the input data. It provides methods for loading and saving the model, calculating p-values, and determining model means based on input size.

​

Code Structure

The SizeMatchedModel class is the central component. It uses a list of bins to partition the input size range and associates a set of parameters with each bin. The distribution attribute holds a statistical distribution object (not defined in this file) used for calculations.

​

Symbols

​

SizeMatchedModel

​

Description

This class represents a statistical model where parameters are determined by the size of the input data. It uses bins to divide the size range and associates a set of parameters with each bin.

​

Inputs

Name	Type	Description
bins	list	List of bin edges defining the size ranges.
params	list	List of parameter sets, one for each bin.
distribution	object	An instance of a statistical distribution class (e.g., from scipy.stats).
name	str	Optional name for the model.

​

Outputs

N/A - This is a class definition, not a function.

​

Internal Logic

The class stores the bins, parameters, distribution, and name. It provides methods for:

• Loading and saving the model from/to JSON files.
• Finding the appropriate parameters for a given size using _params_for_size.
• Calculating p-values using the provided distribution and parameters.
• Determining the model mean for a given size.

​

SizeMatchedModel.from_json

​

Description

This class method loads a SizeMatchedModel instance from a JSON file.

​

Inputs

Name	Type	Description
filename	str	Path to the JSON file containing the model data.

​

Outputs

Name	Type	Description
model	SizeMatchedModel	A new SizeMatchedModel instance loaded from the file.

​

Internal Logic

• Opens the JSON file and loads the data.
• Converts string representations of lists and the distribution name to their respective Python objects.
• Instantiates a new SizeMatchedModel with the loaded data.

​

SizeMatchedModel.to_json

​

Description

This method saves the SizeMatchedModel instance to a JSON file.

​

Inputs

Name	Type	Description
outfile	str	Path to the output JSON file.

​

Outputs

N/A - The method writes data to a file as a side effect.

​

Internal Logic

• Creates a dictionary containing the model’s attributes.
• Converts lists to JSON-serializable strings.
• Extracts the distribution’s class name for serialization.
• Writes the dictionary to the specified JSON file.

​

SizeMatchedModel._params_for_size

​

Description

This private method retrieves the appropriate parameters for a given size based on the defined bins.

​

Inputs

Name	Type	Description
size	float	The size value for which to find the parameters.
strict_bounds	bool	If True, raises an error if the size is outside the defined bins. If False, uses the parameters of the nearest bin.

​

Outputs

Name	Type	Description
params_match	tuple	The parameters associated with the bin containing the given size.

​

Internal Logic

• Uses np.digitize to find the index of the bin corresponding to the input size.
• Handles cases where the size is outside the defined bins based on strict_bounds.
• Adjusts the bin index to match the zero-based indexing of the params list.
• Returns the parameters for the selected bin.

​

SizeMatchedModel.pvalue

​

Description

This method calculates the p-value of a given value x under the model for a specific size.

​

Inputs

Name	Type	Description
x	float	The value for which to calculate the p-value.
size	float	The size value used to determine the model parameters.
invert_cdf	bool	If True, calculates 1 - CDF(x) instead of CDF(x).
strict_bounds	bool	Passed to _params_for_size to control behavior for sizes outside the defined bins.

​

Outputs

Name	Type	Description
p	float	The calculated p-value.

​

Internal Logic

• Retrieves the model parameters for the given size using _params_for_size.
• Calculates the cumulative distribution function (CDF) of x using the model’s distribution and parameters.
• Inverts the CDF if invert_cdf is True.
• Returns the calculated p-value.

​

SizeMatchedModel.model_mean

​

Description

This method calculates the mean of the model for a given size.

​

Inputs

Name	Type	Description
size	float	The size value used to determine the model parameters.
strict_bounds	bool	Passed to _params_for_size to control behavior for sizes outside the defined bins.

​

Outputs

Name	Type	Description
mean	float	The calculated mean of the model.

​

Internal Logic

• Retrieves the model parameters for the given size using _params_for_size.
• Calculates the mean of the distribution using the retrieved parameters.
• Returns the calculated mean.

​

TODOs

• Evaluate distribution in global namespace, so that import is not necessary here