PMF statistics to estimate factor loading

Usage

asap_pmf_regression_cbind_mtx(
  mtx_files,
  row_files,
  col_files,
  idx_files,
  log_beta,
  beta_row_names,
  log_delta = NULL,
  r_batch_names = NULL,
  rename_columns = FALSE,
  do_stdize_beta = TRUE,
  do_stdize_r = TRUE,
  do_log1p = FALSE,
  verbose = FALSE,
  a0 = 1,
  b0 = 1,
  max_iter = 10L,
  NUM_THREADS = 0L,
  BLOCK_SIZE = 1000L,
  MAX_ROW_WORD = 2L,
  ROW_WORD_SEP = "_",
  MAX_COL_WORD = 100L,
  COL_WORD_SEP = "@"
)

Arguments

log_beta: D x K log dictionary/design matrix
beta_row_names: row names log_beta (D vector)
log_delta: D x B log batch effects
r_batch_names: batch names (optional)
rename_columns: append batch name at the end of each column name (default: FALSE)
do_stdize_beta: use standardized log_beta (default: TRUE)
do_stdize_r: standardize correlation matrix R (default: TRUE)
do_log1p: do log(1+y) transformation (default: FALSE)
verbose: verbosity
a0: gamma(a0, b0) (default: 1)
b0: gamma(a0, b0) (default: 1)
max_iter: maximum iterations (default: 10)
NUM_THREADS: number of threads in data reading
BLOCK_SIZE: disk I/O block size (number of columns)
MAX_ROW_WORD: maximum words per line in row_files[i]
ROW_WORD_SEP: word separation character to replace white space
MAX_COL_WORD: maximum words per line in col_files[i]
COL_WORD_SEP: word separation character to replace white space
mtx_file: matrix-market-formatted data file (D x N, bgzip)
row_file: row names file (D x 1)
col_file: column names file (N x 1)
idx_file: matrix-market colum index file

Value

a list that contains:

beta the dictionary matrix (row x factor)
delta the dictionary matrix of batch effects (row x batch)
corr empirical correlation (column x factor)
theta factor loading (column x factor)
log.theta log-scaled factor loading (column x factor)
colsum column sum (column x 1)
rownames row names
batch.names batch names (based on
batch.index
colnames column names