negloglik: negative log-likelihood function of a CARseq negative...
In Sun-lab/CARseq: Cell type-specific differential analysis of count data

negloglik provides a negative log-likelihood function of negative binomial distribution whose mean is a sum of nonnegative terms with covariates

negloglik(
  coefs,
  cell_type_specific_variables,
  other_variables,
  read_depth,
  cellular_proportions,
  counts,
  use_log_scale_params = FALSE,
  verbose = FALSE
)

`coefs`	a vector of c(beta, gamma, overdispersion), where beta is a length K vector of cell type-indepedent coefficients, gamma is a matrix of H x M dimension of cell type-specific non-negative coefficients, and overdispersion is a scalar.
`cell_type_specific_variables`	an array of n_B x H x K of cell type-independent variables.
`other_variables`	a design matrix of n_B x M of cell type-specific variables. Can be NULL.
`read_depth`	a vector of n_B sample-specific read depths.
`cellular_proportions`	a matrix of n_B x H of cellular proportions.
`counts`	A vector of n_B total read counts observed.
`use_log_scale_params`	logical. The default is FALSE, where the cell type-specific effects are parametrized on a non-log scale. Otherwise, they are parametrized on a log scale.
`verbose`	logical. If `TRUE`, a matrix fitted cell type-specific expression will be attached.

library(CARseq)
set.seed(1234)
H = 4
n_B = 60
K = 1
M = 3
overdispersion_theta = 10
coefs = c(rep(0, K), rep(1, H*M), overdispersion_theta)  # initial value
cell_type_specific_variables = array(0, dim=c(n_B, H, M))
cell_type_specific_variables[1:(n_B/3), , 1] = 1
cell_type_specific_variables[(n_B/3+1):(n_B*2/3), , 2] = 1
cell_type_specific_variables[(n_B*2/3+1):n_B, , 3] = 1
other_variables = matrix(runif(n_B * K, min = -1, max = 1), nrow=n_B, ncol=K)
read_depth = round(runif(n_B, min = 50, max = 100))
cellular_proportions = matrix(runif(n_B * H), nrow = n_B, ncol = H)
cellular_proportions = cellular_proportions / rowSums(cellular_proportions)
counts = round(runif(n_B, min = 100, max = 200))
CARseq:::negloglik(coefs, cell_type_specific_variables, other_variables, read_depth, cellular_proportions, counts)