Perform Empirical Bayes Matrix Factorization using flashier, and
return a list of candidate covariance matrices

```
cov_flash(
data,
factors = c("default", "nonneg"),
subset = NULL,
remove_singleton = FALSE,
tag = NULL,
output_model = NULL,
greedy_args = list(),
backfit_args = list()
)
```

## Arguments

- data
A “mash” data object.

- factors
If `factors = "default"`

, the factors and
loadings are both unconstrained. If `factors = "nonneg"`

, the
factors are constrained to be non-negative, and the loadings are
unconstrained.

- subset
Data samples (rows) used to estimate the
covariances. Sset to `NULL`

to use all the data.

- remove_singleton
If `remove_singleton = TRUE`

, factors
corresponding to singleton matrices will be removed from the output.

- tag
How to name the covariance matrices.

- output_model
The fitted flash model will be saved to this file
(using `saveRDS`

).

- greedy_args
List containing additional parameters passed to
`flashier::flash_greedy`

.

- backfit_args
List containing additional parameters passed
to `flashier::flash_backfit`

.

## Value

A list of covariance matrices.

## Examples

```
# See https://stephenslab.github.io/mashr/articles/flash_mash.html
# for an example
```