Distributions

In order to call p-values for individual 5C interactions, we simply compare the observed value of the interaction to its expected value (see Expected modeling) and its variance estimate (see Variance modeling) by using a statistical distribution parameterized to have mean and variance equal to the predictions of the expected and variance models, respectively.

Conceptual overview

We may wish to call p-values using a variety of different statistical distributions, each of which has a unique “native” parameterization. To keep things from becoming too complicated, we ignore these “native” parameterizations and instead always parameterize distributions with the same two parameters: mean and variance.

Command-line interfaces

To call p-values, run

$ lib5c pvalues

To plot the parameterized distributions over the real data, run

$ lib5c plot visualize-fits

Exposed functionality

The exposed function is lib5c.util.distributions.call_pvalues().

Additional utility functions are provided in lib5c.util.distributions to ease parameter conversion, etc.

The following functions are exposed for visualization of distributions:

• lib5c.plotters.fits.plot_fit()

• lib5c.plotters.fits.plot_group_fit()

plot_fit() is a reusable function for overlaying distributions.

plot_group_fit() is a convenience function for overlaying parametrized distributions over groups of real data points.