0
$\begingroup$

I have y data at a daily frequency and a number of x variables at daily, weekly and monthly intervals. I'm looking to create a multiple linear regression from these.

I'm familiar with MIDAS methods, but these typically prescribe using higher frequency x vars with lower freq y variables.

Have considered simply up/downsampling approaches, but averages or linear interpolation is likely to defeat the purpose of attempting the regression in the first place.

There appear to be a few applicable R packages (tempdisagg, tsibbledata) and also found this and this, but wonder if someone more familiar with this kind of work can advise what tends to be de facto aside from trying a bunch of approaches and seeing what works/is feasible. As this seems like a common enough problem wondering if there are common enough approaches to start with.

Working primarily with Python but seems like R may be more appropriate for actual implementation.

Improve this question
$\endgroup$
2
  • $\begingroup$ Your question is a bit borderline for CV. It appears you are looking for the right statistical approach but also which package may be necessary to accomplish that. Programming questions are typically off-topic here unless the query is specifically about the statistics behind the approach (such as an explanation of why certain forecasting methods may be suitable for you), which I reckon your question is more about. I would clarify that or else this question may be closed or moved to Stack Overflow. $\endgroup$
    – Shawn Hemelstrand
    Commented Dec 27, 2023 at 0:48
  • $\begingroup$ @ShawnHemelstrand, concerned just with approach for the time being. Reference to programming/packages is more secondary (and an attempt to show what I've looked at/attempted thus far) $\endgroup$
    – Chris
    Commented Dec 27, 2023 at 22:15

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.