By John Mount
In R analyses are typically organized around a data structure called a data.frame. data.frames are similar to database tables and provide a number of invariants (subject to the usual exceptions you always find in R) in that:
- Data frames are a two dimensional array of cells, addressed by row (representing intances) and columns (representing measurements).
- Cells are addressable by rows indices and column indices/names.
- Every cell in a column has the same type.
- Every column has the same number of rows.
This is a very powerful basis for analysis and software. It is fairly central to R and to teaching R.
As you work with more data you may want to use some higher performance and higher power extensions of data.frame. In particular the packages data.table and dplyr are two powerful high quality possibilities.
However, we do not certify Win-Vector LLC packages ( vtreat, WVPlots ) to safely work with data.table. This is due to a design choice in the otherwise excellent data.table package that we are not comfortable attempting working around. Obviously data.table is a much more influential package than any of ours, but I think the issue is still worth discussing. Our suggested work-around is to convert any data.table you are working with to a native data.frame with an as.data.frame() call on the way into vtreat or WVPlots. Or, as I am sure some readers have already decided, you can just avoid vtreat and WVPlots.
data.frame
data.frame supplies a number of ways to access data cells. The common ones are illustrated below.
Consider the following simple data.frame:
d <- data.frame(x=1:3,y=5:7)
We can get the contents of the column x through a number of different notations:
d[,'x']d[,1]d[,c(TRUE,FALSE)]d$xd[['x']]
Now some of these should have the additional argument drop=FALSE set (which …read more
Source:: win-vector.com