Question

I have a table with 2 columns and I want to plot values from the second column against the sum of N values in the first column in the 5 rows preceding the current row.

How can I achieve this transformation in R without pre-processing externally?

Example:

File: data.csv

7,2
4,8
3,6
7,10
9,3
0,4
8,9
3,4
5,3
4,6

d = read.csv("data.csv", col.names=c("a","b"))
plot(??some_transform??(d$a), d$b)

some_transform should yield the following column:

7 = 7
7+4 = 11
7+4+3 = 14
7+4+3+7 = 21 # Up to this is optional, can start with row 5
7+4+3+7+9 = 30
4+3+7+9+0 = 23
3+7+9+0+8 = 27
7+9+0+8+3 = 27
9+0+8+3+5 = 25
0+8+3+5+4 = 20

So, the plot input is

7,2
11,8
14,6
21,10
30,3
23,4
27,9
27,4
25,3
20,6

Answer 1

The embed function will assemble a matrix with increasing lags from a vector. Take a look at:

 embed(DF[ , ], 5)

So if this is intended to start working at the 5th row, then you could use these results:

 apply(embed(DF[ , 1], 5), 1,sum)
 plot(apply(embed(DF[ , 1], 5), 1,sum) , DF[-(1:4), 2])

EDIT: With your changed specification, add 4 zeros to the beginning of the embed argument. I am displaying the result of using that approach so it is easier to “visualize” the result of that extension to the embed first argument:

> embed( c(0,0,0,0, DF[ ,1 ]), 5)
      [,1] [,2] [,3] [,4] [,5]
 [1,]   80    0    0    0    0
 [2,]   80   80    0    0    0
 [3,]   80   80   80    0    0
 [4,]   81   80   80   80    0
 [5,]   82   81   80   80   80
 [6,]   82   82   81   80   80
 [7,]   83   82   82   81   80
 [8,]   83   83   82   82   81
 [9,]   90   83   83   82   82
[10,]   90   90   83   83   82
[11,]   90   90   90   83   83
[12,]   84   90   90   90   83
[13,]   84   84   90   90   90
[14,]   85   84   84   90   90
[15,]   85   85   84   84   90

(Usually with embed (see further above) one gets a shortened array, but our questioner was willing to accept the “lead-in” phenomenon.)

apply(embed( c(0,0,0,0, DF[ ,1 ]), 5), 1,sum)
 plot( apply(embed( c(0,0,0,0, DF[ ,1 ]), 5), 1,sum), DF[, 2])