how to summarise a numeric variable by two factors in R

Question

I have a data.frame with 3 variables and 1.425.558 observations. It´s a register of installed electric power from renewable energy plants. Every row stands for one installed power plant. There can be multiple power plants of the same type in a zipcode-area.

ID  zipcode     Type    power
1   79280   solarpower  3
2   79280   solarpower  3
3   79283   hydroelectric   3
4   79280   biogas          55
5   79280   windpower   2
6   21459   windpower   4
7   21459   windpower   2

I would like to sum by zipcode how much solarpower/biogas/windpower is installed.

zipcode     Type    power
21459        windpower    6
79280        solarpower   6
79280        windpower    2
...and so on.

I already tried

aggregate(myDat$power, by=list(myDat$zipcode,myDat$type), FUN=sum)

but my RAM was not sufficant.

I know, my dataframe is very big. I could narrow it down a lot, because i only need the data for those zipcodes that start with "2".

Could you point me to a solution? Thank you very much for helping an Beginner!

Answer 1

If I understand correctly what you need, you can express it using dplyr:

> data %.% group_by( zipcode, Type ) %.% summarise( power = sum(power) )
Source: local data frame [5 x 3]
Groups: zipcode

  zipcode          Type power
1   21459     windpower     6
2   79280     windpower     2
3   79280        biogas    55
4   79283 hydroelectric     3
5   79280    solarpower     6

And if you only want those zip code that start by 2, you can filter first:

> data %.% filter( grepl( "^2", zipcode ) ) %.% 
     group_by( zipcode, Type ) %.% summarise( power = sum(power) )
Source: local data frame [1 x 3]
Groups: zipcode

  zipcode      Type power
1   21459 windpower     6

Answer 2

data.table version:

library(data.table)
dt = data.table(your_df)

dt[, sum(power), by = list(zipcode, Type)]

And to narrow down first:

dt[grep("^2", zipcode), sum(power), by = list(zipcode, Type)]

Because grep is expensive, in both dplyr and data.table you're likely better off (speed-wise) summarizing first, and filtering second, i.e.:

dt[, sum(power), by = list(zipcode, Type)][grep("^2", zipcode)]