1

I have consulted this question on the basis of prediction intervals for loess (How to calculate prediction intervals for LOESS?). However, I am unaware of whether it is proper or possible to use prediction intervals for forecasts made from loess? What problems could come from doing so, if any? How would I do so in R?

Currently I have attempted to calculate forecasts by using a for loop that predicts the values for a certain number of subsequent days, but I am unaware of how to produce prediction intervals for the same set of forecasts. I would hope to adjust the forecast so that my time-series and prediction look similar to this:

enter image description here

Community
  • 1
a.powell
  • 1,047
  • 7
  • 19
  • Extrapolation from a LOESS is a big *no-no*... – usεr11852 Jul 26 '16 at 00:51
  • @usεr11852 is this because too the weight is placed on locally? What other methods would you suggest for modeling data in which I want to extrapolate but also weight more recent observations? – a.powell Jul 26 '16 at 01:00

0 Answers0