How to know the type of missing data is what it is: MCAR, MAR or NMAR, knowing that I'm working on time series multivariate, and is that going to help me deal with the missing data, and what is the best techniques of processing missing data in time series, knowing that I work on meteorological data?
Topic time-series
Category Data Science
It is a question related to the domain of your project. You should know the cause of the missingness.
If some values are missing, because there is no applicable measure then this values are a special case, therefore they're missing not at random. In such case you can impute missing values with the mean value of the non-missing data and add another feature, which indicates special cases.
On the other hand, if the value is missing, because some sensor temporarily stopped working, then it is missing at random, so the measured value is just not known. In this situation you can perform linear regression (or any other regression, but you should start with the linear one) between non-missing data.
In case your problem has mixed types of missingness you should also perform some regression.
Geeks Mental is a community that publishes articles and tutorials about Web, Android, Data Science, new techniques and Linux security.