Model selection over candidate latent dimensions (BIC or ICL)

For each value in latents, fits a model with fun and computes the chosen criterion (BIC or ICL). Returns the criterion values, the latent dimension that maximizes the criterion and the corresponding fit.

Usage

select_model_bic(
  Y,
  X,
  fun,
  latents,
  model = c("ZIPLNPCA", "PLNPCA"),
  criterion = c("BIC", "ICL")
)

Arguments

Y: Numeric n x p count matrix (NAs allowed).
X: Design matrix, (n*p) x d.
fun: Fitting function, called as fun(Y = Y, X = X, q = q) (e.g. Miss.ZIPLNPCA or Miss.PLNPCA).
latents: Integer vector of candidate latent dimensions.
model: "ZIPLNPCA" (default) or "PLNPCA"; must match fun. Passed to BIC and ICL.
criterion: "BIC" (default) or "ICL".

Value

A list with elements:

criterion: the criterion used.
values: numeric vector of criterion values, named by latents.
q_star: the latent dimension maximizing the criterion.
fit_star: the fit obtained with q_star.

Details

A progress bar is displayed via pbapply. The function checks that model matches the fits returned by fun: a ZIP-PLN-PCA fit contains mStep$gamma, a PLN-PCA fit does not.

Examples

set.seed(1)
n <- 30; p <- 5; d <- 2; q <- 2
X <- cbind(1, rnorm(n * p))                       # (n*p) x d design
theta <- list(B = matrix(c(1, 0.5)), D = matrix(c(1, -0.5)),
              C = matrix(rnorm(p * q, sd = 0.5), p, q))
Y <- Simul(X, theta, list(n = n, p = p, d = d, q = q))$Y
Y[sample(n * p, 20)] <- NA                        # missing entries

# \donttest{
sel <- select_model_bic(Y, X, Miss.ZIPLNPCA, latents = 1:3)
sel$values
#>         1         2         3 
#> -265.5805 -267.4293 -272.9659 
sel$q_star
#> [1] 1
fit <- sel$fit_star

sel_icl <- select_model_bic(Y, X, Miss.ZIPLNPCA, latents = 1:3,
                            criterion = "ICL")
sel_icl$q_star
#> [1] 1
# }