0

I've encountered a little problem with my research. My study is about the relationship between ICT Development and Service Trade from 1999-2018 (annually). Some of the missing data are located in the middle. Therefore I use Ipolate command in Stata. But there are missing data from 1999-2000 and some of them are in the last period (2017-2018) and the Ipolate command didn't work. What can I do to deal with this? Should I let the Stata work with the missing data or should I predict them? Also, what is the maximum number of observations that are allowed to be interpolated/extrapolated in panel data?

Thanks a lot.

Coraline
  • 1

1 Answers1

1

First of all interpolating, or extrapolating, is a kind of naive approach to predicting time-series values. I call it naive, because you assume the linear relationship that is used for interpolation, that in many cases won't hold. In general, you should never blindly use interpolation in such cases, since it may often lead to bad results. It might be ok to use interpolation for values that "don't change that much" over time, while you have only small gaps to fill, but it might be very bad for other data. I recommend you the What is wrong with extrapolation? thread for many great examples and explanations.

Tim
  • 108,699
  • 20
  • 212
  • 390