Question

I am trying to figure out a method to increase the number of elements of vectors, in order to remove the angular effect visible when I plot the values of these vectors. For instance, let’s say I got two vectors containing 10 elements each:

a = c(4,2,10,5,3,4,8,9,6,2)
b = seq(0,4.5, by=0.5)

They are subject to data smoothing, but I would like to increase their “resolution” to obtain more prediction points than 10 (its length). So, in other words, take the vector a and double (for example) its number of elements, while keeping its consistency. The resulting vector should be something like:

a = c(4,3,2,6,10,7,5,4,3,4,6,8,8.5,9,7.5,6,4,2,2)

Of course, in this particular case, I can easily compute the average of the elements pair-wise. But I would like to have a generalized method for an arbitrary length. I have tried with:

seq(a[1],a[10], length.out=20)

but of course this does not do the job as only the first and last element of the vector are taken in consideration. It is suitable for the second vector b though (which contains the abscissa values).
Any help would be appreciated. Thanks.
Marius.

Answer 1

Do you just want linear interpolation based on the b sequence?

 f <- approxfun(b, a)
 f(seq(b[1], b[length(b)], length = 20))
 [1] 4.000000 3.052632 2.105263 5.368421 9.157895 8.157895 5.789474 4.368421 3.421053 3.263158 3.736842 4.842105 6.736842 8.157895 8.631579 8.684211 7.263158 5.789474
 [19] 3.894737 2.000000

See ?approx for other options and its See also for other forms of interpolation, also ?round for rounding out the values in various ways.