# [R] Question re: interpolation

David Winsemius dwinsemius at comcast.net
Sun May 2 23:44:03 CEST 2010

```On May 2, 2010, at 5:27 PM, Jessica Schedlbauer wrote:

> Hello,
>
> I have a matrix in which two variables, x and y, are used together
> to determine z.  The variables x and y are sorted into classes.
> Specifically, values for variable x range from 0 to 2.7 and are
> sorted into class increments of 0.15 and variable y ranges from
> 0-2100 with class increments of 100.  For each x-y class
> combination, I have calculated a mean value of z from existing
> data.  However there are gaps in the existing data (i.e., x-y
> combinations for which there are no z data).  I would like to
> interpolate values to fill in these missing datapoints, but have so
> far been unable to find a straightforward way of doing so in R.
> There are many gaps in the data and I need to repeat this process
> for several individual datasets.  That said, I am looking for a way
> to avoid interpolating each missing point separately.

Why not do a loess fit _before_ you do any binning and then
predict.loess at the intervals where you want estimates?

x=rnorm(200, 1.5)
y= (10*x)^2 +rnorm(200, 0, 100)
plot(x,y, cex=0.5, xlim=c(-2, 5))
points(seq(-2,5, by=0.15), predict(loess(y~x),
newdata=data.frame(x=seq(-2,5, by=0.15))),
col="red", cex=2)
>
> Any suggestions would be greatly appreciated.
>
> Regards,
> Jessica Schedlbauer
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> and provide commented, minimal, self-contained, reproducible code.

David Winsemius, MD
West Hartford, CT

```