Unfortunately, time series usually ave missing values, which can be caused by a variety of reasons. Time series data, due to its nature (one observation per data point), could become useless, in principle, because of a single missing value. For monthly and yearly data, missing values are an unfortunate reality that occurs frequently.
Imputing missing values for standard data (non-time series) is usually easy: the average or median is usually used without causing much trouble. In a time series context, we need to take much more care:
- Time series usually have some seasonality, so the imputation should take that into consideration
- In time series we usually have few observations, so any value that is incorrectly imputed can have serious consequences on the overall estimation
