geometric interpolation - missing values

[Vicente Neira]

Dear Statlist users,

I want to fill missing values in time-series with geometrical interpolation (i.e., the growth rates are the same between the interpoled values).
I have been exploring all the *polate options (ipolate, csipolate, etc), and the multiple imputation options, but I thinks none of them can do geometric interpolation.
It's very strange, since it's a standard procedure with censal data.
Thanks,

Vicente.

[William Lisowski]

Welcome to Statalist!

I hope someone else can answer this definitively, but I think that linear interpolation on the logarithms of your time series values would accomplish geometric interpolation as you define it.

Some work with Google searching "geometric interpolation census data" and similar phrases has found nothing useful in the way of even a basic "how-to" on geometric interpolation as used in this context.

[Nick Cox]

I haven't searched for the post but I am pretty sure this has been asked before.

William's answer of take logarithms, interpolate on the logarithms using ipolate and then exponentiate is certainly precisely what springs to mind.

[Vicente Neira]

Yes, that do the trick, thanks a lot.