shrinkDataFrame: Shrink data.frame by row groups

In jmw86069/jamses: Jam SummarizedExperiment Stats

Shrink data.frame by row groups

Description

Shrink data.frame by row groups

Usage

shrinkDataFrame(
  x,
  groupBy,
  na.rm = TRUE,
  string_func = function(x) jamba::cPasteSU(x, na.rm = TRUE),
  num_func = function(x) {
     mean(x, na.rm = TRUE)
 },
  add_string_cols = NULL,
  num_to_string_func = as.character,
  keep_na_groups = TRUE,
  include_num_reps = FALSE,
  collapse_method = 2,
  verbose = FALSE,
  ...
)

Arguments

x	data.frame (or equivalent)
groupBy	character vector with one of the following: one or more columns in colnames(x). The values in these columns will define the row groups used. character or factor with length equal to nrow(x). These values will define the row groups used.
string_func	function, default uses jamba::cPasteSU(), used for character or factor columns. Note that string columns are handled differently than numeric columns by applying vectorized operations across the complete set of rows in one step, rather than calling data.table on each subgroup.
num_func	function, default function(x)mean(x, na.rm=TRUE), used for numeric columns. Note that this function is applied to each row group by data.table, and is typically very efficient for numeric values.
add_string_cols	character with optional numeric columns that should be handled as if they were character columns. Default NULL.
num_to_string_func	function used for add_string_cols when converting numeric columns to character. Default as.character() retains the full numeric value, however it may be useful to use something like function(x)signif(x, digits=3) to limit the output to only three significant digits, or function(x)format(x, digits=3).
keep_na_groups	logical, default TRUE, whether to convert NA values in row groups to "" so they are retained in the output. You may want to use keep_na_groups=FALSE when there are a large number of un-annotated rows that should not be aggregated together. This situation may occur if converting a probe to a gene symbol, where a subset of probes cannot be converted to a gene symbol and instead receive NA.
include_num_reps	logical indicating whether to add a column "num_reps" to the output, with the integer number of rows in each row group.
collapse_method	integer default 2, indicating the internal collapse method used. Experimental. 1 collapses each numeric column independently. 2 collapses each set of numeric columns that use the same numeric shrink function. When all numeric columns use the same shrink function, they are all calculated in a single step, which is typically much faster.
verbose	logical indicating whether to print verbose output.
...	additional arguments are ignored.

Details

Purpose is to shrink a data.frame to have one row per row grouping. The row grouping can use a single column of identifiers, or multiple columns. The challenge is to apply a relevant function to each column, expecting there will be columns with numeric, character, or factor types.

The default behavior:

• numeric columns are summarized with mean(x, na.rm=TRUE), so that NA values are ignored when there are non-NA values present.

• character columns are combined using unique, sorted character strings.

  • This step uses jamba::cPasteSU() where the S activates sorting using jamba::mixedSort(), and U calls unique().

  • To retain all values, remove the U and call jamba::cPasteS()

  • To skip the sort, remove the S and call jamba::cPasteU()

  • To keep all values, and skip sorting, call jamba::cPaste()

See Also

Other jamses utilities: choose_annotation_colnames(), contrast2comp_dev(), fold_to_log2fold(), intercalate(), list2im_opt(), log2fold_to_fold(), make_block_arrow_polygon(), mark_stat_hits(), matrix_normalize(), point_handedness(), point_slope_intercept(), shortest_unique_abbreviation(), shrink_df(), shrink_matrix(), sort_samples(), strsplitOrdered(), sub_split_vector(), update_function_params(), update_list_elements()

Examples

testdf <- data.frame(check.names=FALSE,
   SYMBOL=rep(c("ACTB", "GAPDH", "PPIA"), c(2, 3, 1)),
   `logFC B-A`=c(1.4, 1.4, 2.3, NA, 2.5, 5.1),
   probe=paste0("probe", 1:6))
shrink_df(testdf, by="SYMBOL")

shrink_df(testdf, by="SYMBOL", num_func=mean)

shrink_df(testdf, by="SYMBOL", add_string_cols="logFC B-A")

testdftall <- do.call(rbind, lapply(1:10000, function(i){
   idf <- testdf;
   idf$SYMBOL <- paste0(idf$SYMBOL, "_", i);
   idf;
}))
shrunk_tall <- shrink_df(testdftall,
   by="SYMBOL")
head(shrunk_tall, 6)

shrunk_tall2 <- jamses::shrinkDataFrame(testdftall,
   groupBy="SYMBOL")
head(shrunk_tall2, 6)

jmw86069/jamses documentation built on Nov. 4, 2024, 9:25 p.m.