Hi
I have a collection of about 16 time series with occasional missing
data. A few of these time-series start later than the rest. There is a
relatively high correlation between them (they are hourly temps at
various locations around the UK). The longest series contains about
40000 points
I have constructed each time series into a zoo object and then merged
each of these objects in to one (zoo is great!). Using na.approx I can
interpolate through a single column for the occasional missing data. Now
I would like to cross interpolate those series which start later than
the rest.
I thought of reducing the collated zoo object down to a time span where
all columns where defined, then calculating a correlation matrix. Then
taking this matrix back to the larger collated zoo object and then
interpolating the missing values based on the correlations I have found.
I'm assuming correlation is stationary year to year.
Has anybody tried anything like this? Is a function out there that could
help?
Thanks
Sean
[[alternative HTML version deleted]]