I have weekly data of one of the columns of my dataset and all other data are collected on a daily basis. Is there any way to convert that weekly data to daily data?
While it is always possible to go from higher frequency observations to lower frequency observations by means of any process of aggregation, the converse is generally difficult and not always feasible and it will depend on what data and variables you are dealing with. Made this premise, from my knowledge, one method to accomplish the task is using cubic spline interpolation, if you do a quick research you will find something. However, this will introduce a series of problems.
Personally, I would suggest you to keep the weekly data and convert the other daily data into weekly data, which is always possible and less problematic. Alternatively, you can consider the class of MIDAS models which are conceived to deal with different-sampled data. Hope this will help you!
As Giuseppe says, this is not easy. It is difficult to give an answer when we don't know what the data is, or what it is being used for. If the data is the output of a system to which you are calibrating a model, keep the model inputs at a daily time step, and aggregate the model outputs to a weekly time step and then compare with the observations.
Otherwise, you can disaggregate from weekly to daily using an interpolation method only if the quantity does not vary significantly over time scales shorter than a week. If the quantity varies significantly at a daily time scale, then the only possibility is a stochastic approach, producing a large number of possible daily datasets. This would require knowledge of the statistical properties at a daily time scale. Without this, you should aggregate the daily data to weekly.
Hi,
Please you can watch videos on this using Eviews from Youtube. I think it will be of help.
Thank you and stay safe
- Badges
- Science topic
- Similar topics
- Statistics
- Survey