I have two data.tables showing temperatures for multiple locations (identified by geocode).
I would like to create a third one based on a subtraction of them. Here they are:
library(data.table)
# Generate random data:
geocode <- paste0("N", 1:10)
dates <- seq(as.Date("2000-01-01"), as.Date("2004-12-31"), by="month")
models <- c("A", "B", "C", "D", "E")
temp <- runif(length(geocode)*length(dates)*length(models), min=0, max=30)
dt1 <- data.table(expand.grid(Location=geocode,Date=dates,Model=models),Temperature=temp)
ref <- runif(length(geocode), min=0, max=30)
dt2 <- data.table(expand.grid(Location=geocode), Temperature=ref)
I would like to conditionally subtract dt2 from dt1. By each location (geocode), I would like to subtract temperature in dt2 from temperature in dt1, preserving the other columns (Date and Model).
How to achieve this? I would know how to do if it was a single data table, but I've never tried to do algebra on two different data tables like this before.
I think this works:
dt1[dt2, on=.(Location), td := x.Temperature - i.Temperature, by=.EACHI]
Location Date Model Temperature td
1: N1 2000-01-01 A 3.949276 -19.2110455
2: N2 2000-01-01 A 2.811684 -11.6405195
3: N3 2000-01-01 A 24.069659 13.6159779
4: N4 2000-01-01 A 25.809426 -1.8793405
5: N5 2000-01-01 A 25.193624 19.6812965
---
2996: N6 2004-12-01 E 24.298463 4.0218859
2997: N7 2004-12-01 E 1.488011 -26.4472283
2998: N8 2004-12-01 E 27.489108 5.6525076
2999: N9 2004-12-01 E 3.487664 -5.9926003
3000: N10 2004-12-01 E 8.523718 -0.7559126
Checking by eye...
dt2[dt1[1:5], on=.(Location), .(Location, t1 = i.Temperature, t2 = x.Temperature)]
Location t1 t2
1: N1 3.949276 23.160321
2: N2 2.811684 14.452204
3: N3 24.069659 10.453681
4: N4 25.809426 27.688766
5: N5 25.193624 5.512328
Looks right to me.
How it works
The syntax for an update join is x[i, v := expr, by=.EACHI]. Inside the expression, prefixes i.* and x.* can be used to clarify where columns are being taken from.
The by=.EACHI might not be needed, but I usually use it for this.