Question

I am trying to solve this in R, although I know I would have done it by now in Excel, I really want to learn how to use R.

My dataframe looks like this:

 OBJECTID CDUID SENSI_FINA FREQUENCY SUM_LENGTH

        6  5915          1        51      19178

        7  5915          2        97      21536

        8  5915          3       201      35640

        9  5915          4       551     170549

       10  5915          5       308     145126

       11  5917          1       210      28104

       12  5917          2      1897     249379

Now I would like to sum the SUM_LENGTH per CDUID and then calculate the percentage the SUM_LENGTH with SENSI_FINAL=5 is of the summed SUM_LENGTH per CDUID

So in simple words I want to do this:

(145123/(19178+21536+35640+170549))*100

for CDUID = 5915 and then for the next 5917 etc..

What I did so far is I calculated the sum based on the CDUID:

CDlenght <- aggregate(step1$SUM_LENGTH~step1$CDUID, data=step1, sum)

but now I’m stuck…:-(

Answer 1

I would use data.table or ddply for this

library(plyr)

DT2 <- ddply(step1, .(CDUID), mutate, percent = 100* SUM_LENGTH / sum(SUM_LENGTH))

 subset(DT2, SENSI_FINA == 5)
##   OBJECTID CDUID SENSI_FINA FREQUENCY SUM_LENGTH percent
## 5       10  5915          5       308     145126 37.0192

I think the data.table syntax is more elegant, and it is more memory efficient

library(data.table)
DT <- data.table(step1)

DT[,percent :=  100* SUM_LENGTH / sum(SUM_LENGTH), by = CDUID][SENSI_FINA == 5]

:= assigns by reference within DT (so the percent column will now be in DT