[R] select rows by criteria

[Petr Savicky]

On Thu, Mar 01, 2012 at 05:42:48PM +0100, Petr Savicky wrote:
> On Thu, Mar 01, 2012 at 04:27:45AM -0800, syrvn wrote:
> > Hello,
> > 
> > I am stuck with selecting the right rows from a data frame. I think the
> > problem is rather how to select them
> > then how to implement the R code.
> > 
> > Consider the following data frame:
> > 
> > df <- data.frame(ID = c(1,2,3,4,5,6,7,8,9,10), value =
> > c(34,12,23,25,34,42,48,29,30,27))
> > 
> > What I want to achieve is to select 7 rows (values) so that the mean value
> > of those rows are closest
> > to the value of 35 and the remaining 3 rows (values) are closest to 45.
> > However, each value is only
> > allowed to be sampled once!
> 
> Hi.
> 
> If some 3 rows have mean close to 45, then they have sum close
> to 3*45, so the remaining 7 rows have sum close to
> 
>   sum(df$value) - 3*45 # [1] 169
> 
> and they have mean close to 169/7 = 24.14286. In other words,
> the two criteria cannot be optimized together.
> 
> For this reason, let me choose the criterion on 3 rows.
> The closest solution may be found as follows.
> 
>   # generate all triples and compute their means
>   tripleMeans <- colMeans(combn(df$value, 3))
> 
>   # select the index of the triple with mean closest to 35
>   indClosest <- which.min(abs(tripleMeans - 35))

I am sorry. There should be 45 and not 35.

  indClosest <- which.min(abs(tripleMeans - 45))

  # generate the indices, which form the closest triple in df$value
  tripleInd <- combn(1:length(df$value), 3)[, indClosest]
  tripleInd # [1] 1 6 7

  # check the mean of the triple
  mean(df$value[tripleInd]) # [1] 41.33333

Petr Savicky.