What are "grouping factors"?
Grouping and "split-apply-combine" explained
The concept of "grouping factors" is widely used in R, allowing
us to quickly split datasets (vectors or data.frames) into subgroups,
work with the subgroups independent (apply functions to them),
and then recombine them as needed.
Various R functions specify "grouping factors" in a confusing
variety of subtly different ways, usually as function arguments named
things like INDEX, INDICES, f, by, or groupby.
In humdrumR, we adopt the tidyverse dplyr approach, using the
group_by() function (and/or the .by argument).
Any atomic vector with at least two unique values, or "levels", can be used as a grouping factor---generally, grouping vectors are coerced into factors. Each unique level in a grouping vector/factor represents a single group. Any vector, or data.frame that is the same length/height as the grouping factor can then be broken into these groups, taking all the indices where the grouping factor equals each group in turn. Since we generally try to work with data.frames, which by definition contain a bunch of vectors that are the same length, we can use any vector/column in a data.frame to group any of the other vectors, or the rows of the whole data.frame.
Most functions allow you to specifiy multiple grouping factors/vectors (so long as they are all
the same length).
The groups are then defined by every unique combination of elements in the vectors.
So, for example, if we use the vectors c('A', 'A', 'A', 'B', 'B', 'B') and
c(1, 1, 2, 2, 3, 3) as grouping factors, we'll get four groups with levels 1A, 2A, 2B, and 3B.
Note that groups created by grouping factors are not neccessarily contiguous.
If we use a vector like c(1, 1, 2, 2, 1, 1) as grouping factor, we get two groups: 1 and 2.
The 1 group would include the 1st, 2nd, 5th, and 6th indices, even though they are
separated in the grouping factor.
If you want contiguous groups you must make them.
The humdrumR function segments() can be used to generate strictly contiguous grouping factors.
For example, segments(c(1, 1, 2, 2, 1, 1)) will return c(1, 1, 2, 2, 3, 3).
See also
Other R lessons.:
evaluatingExpressions,
partialMatching,
recycling,
vectorization