Casting data.tables

Joining Data with data.table in R

Scott Ritchie

Postdoctoral Researcher in Systems Genomics

Casting a long data.table

sales_wide <- dcast(sales_long, quarter ~ year, value.var = "amount")

The dcast() function

The general form of dcast():

dcast(DT, ids ~ group, value.var = "values")
      |   |     |                   |
      |   |     |                    --> column to split
      |   |      ----------------------> group labels to split by
      |    ----------------------------> rows to keep behind as identifiers
       --------------------------------> data.table to reshape

The dcast() function

sales_wide <- dcast(sales_long, quarter ~ year, value.var = "amount")

Splitting multiple value columns

dcast(profit_long, quarter ~ year, value.var = c("revenue", "profit"))

Multiple row identifiers

Keep multiple columns as row identifiers:

dcast(sales_long, quarter + season ~ year, value.var = "amount")

Dropping columns

Only columns included in the formula or value.var will be in the result:

sales_wide <- dcast(sales_long, quarter ~ year, value.var = "amount")

Multiple groupings

Split on multiple group columns:

dcast(sales_long, quarter ~ department + year, value.var = "amount")

Converting to a matrix

sales_wide <- dcast(sales_long, season ~ year, value.var = "amount")
sales_wide

   season    2015    2016
1: Autumn 3420000 3670000
2: Spring 2950000 3000300
3: Summer 2980700 3120200
4: Winter 3200100 3350000

Converting to a matrix

as.matrix() can take one of the columns to use as the matrix rownames:

mat <- as.matrix(sales_wide, rownames = "season")
mat

       2015    2016        
Autumn 3420000 3670000
Spring 2950000 3000300
Summer 2980700 3120200
Winter 3200100 3350000

Let's practice!

Joining Data with data.table in R