guf (version 1.0.2)

aggro: Aggregates a numerical column

Description

The function aggro() does roughly the same as the function aggregate(). However, the differences are that (a) aggro() returns the results ordered by all grouping columns; (b) aggro() is polymorphic, that is, two levels of complexity can be specified; and (c) aggro also displays the results per NA if NAs are values in the grouping column(s) or split.by column.

Usage

aggro(x, group.by, num.column, FUN, split.by = NULL)

Arguments

x

A data.frame with at least one column that can be treated as 'factor', and one numerical column.

group.by

A vector of column names in x. It denotes groups, sub-groups, sub-sub-groups (etc., depending on the number of columns specified) by which counts need to be grouped. See examples.

num.column

A string, denoting the name of the numerical column in x on which to apply function FUN.

FUN

A function that returns a single numerical value when applied to a numerical vector; examples are sum, mean, sd, se, var, etc.

split.by

An optional column name in x, by which to split the counts 'horizontally'. That is, whereas 'group.by' is returned as rows, 'split.by' is returned as columns, whereby every value in 'split.by' will become a column. In that sense, it acts as a pivot specifier. This is one of the main differentiators with aggregate().

Value

A data.frame. The first column(s) is/are the 'group.by' column(s). If 'split.by' is not provided, then the final column name is the column name of the 'num.column'; if 'split'by' is provided, then the return data.frame consists of the columns specified by 'group.by' and the unique values in 'split.by'. See examples.

Examples

# NOT RUN {
my_DF = data.frame(var1=factor(c(rep('low', 4),rep('medium', 4),rep('high', 4)),
levels=c('low', 'medium', 'high')), var2=c(1, 2,2, 3,3,3, 4,4,4,4, 3, 2),
var3=rep(c('bbb','aaa','bbb'), 4), stringsAsFactors=FALSE)
aggro(my_DF, c('var3', 'var1'), 'var2', sum)
# Just like with the function count(), the results are grouped by unique combinations of ...
# ...'var3' and 'var1'. Note the following:
# * Column names are given in parenthesis (either single, ', or double, ")
# * Functions are not specified with parenthesis
# * The output is ordered; in principle according to alphanumerical order, except..
#   ... when a 'group.by' column is an ordered factor, the factor order is followed.
# However, that said, up to this point, the results are the same as with aggregate():
aggregate(var2~var3+var1, my_DF, sum)
# Yet, it is getting more interesting/useful when either the results are split, ...
# ... or when there are NAs envolved; in both cases, aggro() digresses from aggegate();
# see the following:

# With split.by. Also non-factors can be used for 'split.by':
aggro(my_DF, group.by='var1', num.column='var2', FUN=sum, split.by='var3')

# With NAs. For the 'group.by' variable, NAs are treated as 'factor'.
# When there are NAs in the 'split.by' column, then an extra NA column is returned, ...
# ...specifying the counts of the NAs:
my_DF_w_NA = my_DF # same as above, but now...
my_DF_w_NA$var1[1] <- NA
my_DF_w_NA$var2[c(6,10)] <- NA
my_DF_w_NA$var3[10] <- NA
aggro(my_DF_w_NA, c('var1', 'var3'), 'var2', sum)
# Compare with:
aggregate(var2~var1+var3, my_DF_w_NA, sum)

# And indeed, with a split.by:
my_DF_w_NA$var3[8] <- NA
aggro(my_DF_w_NA, group.by='var1', num.column='var2', FUN=sum, split.by='var3')

# }