How to get column mean for specific rows only?

Question

I need to get the mean of one column (here: score) for specific rows (here: years). Specifically, I would like to know the average score for three periods:

• per

Answer 1

If the rows are ordered by year, I think the easiest way to accomplish this would be:

m80_83 <- mean(dataframe[1:4,3]) #Finds the mean of the values of column 3 for rows 1 through 4
m84_90 <- mean(dataframe[5:10,3])
#etc.

If the rows are not ordered by year, I would use tapply like this.

list.of.means <- c(tapply(dataframe$score, cut(dataframe$year, c(0,1983.5, 1990.5, 3000)), mean)

Here, tapply takes three parameters:

First, the data you want to do stuff with (in this case, datafram$score).

Second, a function that cuts that data up into groups. In this case, it will cut the data into three groups based on the dataframe$year values. Group 1 will include all rows with dataframe$year values from 0 to 1983.5, Group 2 will include all rows with dataframe$year values from 1983.5 to 1990.5, and Group 3 will include all rows with dataframe$year values from 1983.5 to 3000.

Third, a function that is applied to each group. This function will apply to the data you selected as your first parameter.

So, list.of.means should be a list of the 3 values you are looking for.

Answer 2

datfrm$mean <-
  with (datfrm, ave( score, findInterval(year, c(-Inf, 1984, 1991, Inf)), FUN= mean) )

The title question is a bit different than the real question and would be answered by using logical indexing. If one wanted only the mean for a particular subset say year >= 1984 & year <= 1990 it would be done via:

mn84_90 <- with(datfrm, mean(score[year >= 1984 & year <= 1990]) )

Answer 3

Since findInterval requires year to be sorted (as it is in your example) I'd be tempted to use cut in case it isn't sorted [proved wrong, thanks @DWin]. For completeness the data.table equivalent (scales for large data) is :

require(data.table)
DT = as.data.table(DF)   # or just start with a data.table in the first place

DT[, mean:=mean(score), by=cut(year,c(-Inf,1984,1991,Inf))]

or findInterval is likely faster as DWin used :

DT[, mean:=mean(score), by=findInterval(year,c(-Inf,1984,1991,Inf))]