Filtering in a time series with gaps due to entity change and missing data

[lal mohan kumar]

Dear all
Sorry for posting the same question again. I couldn't elicit any response, could be due to improper wording or ambiguities.
The case is that I have a micro panel(No: of firms=3000 and no: of years=8). My variable of interest is Cashflows(unscaled) & I am interested in finding trend and cycle components of cashflows. Since it is obvious there will be gaps in Cashflows due to missing data, I am unable to use filtering methods like HP/BK. Below gives the ts report and due to gaps, I am unable to proceed.

Panel variable: id
Time variable: year
---------------------------
Starting period = 2011
Ending period = 2018
Observations = 22256
Number of gaps = 5945
(Gap count includes panel changes).
Thus for each entity, say X, at the best I will have 8 observations(8 Cash flow units) and there will be lots of missing. Also, change in the entity is a gap for time series. In such a case how can I do filtering with cash flows and does it make sense?
In the above case, is there any way to do filtering rather than abandoning the study.

[Nick Cox]

Please note the detailed advice in #1 of https://www.statalist.org/forums/help#adviceextras on practice here.

If a question doesn't attract any attention, the best way to push forward is to post again in the same thread trying to explain more. Posting like this with no cross-reference to your previous thread, and leaving the previous thread dangling, with a risk of replies in both threads, is not a good solution.

I read your previous and it was not clear enough to engage my interest. Here is how it and this too are unclear:

1. What I get is that you want to filter your data and in my case HP/BK triggers a memory of Hodrick-Prescott but at a guess about this community few people here would recognise that. BK is something similar, I presume. Oddly enough, only about 1 of the top 20 posters here is obviously an economist so you are relying on foreign aid here and need to explain what you are doing!

2. I can't see any code here beyond an allusion to tsreport . What filtering code did you try and what report did you get? That would help readers see what you are doing.

Panel lengths of 8 years allow, for what they are worth, fits of e.g. linear and quadratic regression and gaps need not be fatal there. What kinds of cycle are you talking about?