R: Likelihood with data squashing & zero counts

negLLzeroSquash {openEBGM}

R Documentation

Likelihood with data squashing & zero counts

Description

negLLzeroSquash computes the negative log-likelihood based on the unconditional marginal distribution of N (DuMouchel et al. 2001). This function is minimized to estimate the hyperparameters of the prior distribution. Use this function if including zero counts and using data squashing. Generally this function is not recommended unless convergence issues occur without zero counts (negLLsquash is typically more efficient).

Usage

negLLzeroSquash(theta, ni, ei, wi)

Arguments

`theta`	A numeric vector of hyperparameters ordered as: `\alpha_1, \beta_1, \alpha_2, \beta_2, P`.
`ni`	A whole number vector of squashed actual counts from `squashData`.
`ei`	A numeric vector of squashed expected counts from `squashData`.
`wi`	A whole number vector of bin weights from `squashData`.

Details

The marginal distribution of the counts, N, is a mixture of two negative binomial distributions. The hyperparameters for the prior distribution (mixture of gammas) are estimated by optimizing the likelihood equation from this marginal distribution.

The hyperparameters are:

\alpha_1, \beta_1: Parameters of the first component of the marginal distribution of the counts (also the prior distribution)
\alpha_2, \beta_2: Parameters of the second component
P: Mixture fraction

This function will not need to be called directly if using exploreHypers or autoHyper.

Value

A scalar negative log-likelihood value.

Warnings

Make sure ni actually contains zeroes before using this function. You should have used the zeroes = TRUE option when calling the processRaw function.

Make sure the data were actually squashed (see squashData) before using this function.

References

DuMouchel W, Pregibon D (2001). "Empirical Bayes Screening for Multi-item Associations." In Proceedings of the Seventh ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD '01, pp. 67-76. ACM, New York, NY, USA. ISBN 1-58113-391-X.