I have data with big missing gaps over the same months. Should I split this into separate dataframes? Impute over the gaps (seems like a bad idea). Any other ideas? Real lost and am having trouble finding a similar example online to my surprise. I am currently down sampling the readings to weekly mean readings (as some days have multiple observations) and am then splitting this into three separate dataframes... not sure if this is right and if it is where to go from here.
$\begingroup$
$\endgroup$
1
-
1$\begingroup$ carma_pack (python code) fits a CARMA model (continuous-time autoregressive moving average) to arbitrarily sampled time series. You could then use it for forecasting. $\endgroup$– corey979Commented Feb 6, 2020 at 18:08
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
0
This problem arises quite naturally with predicting beer sales where the beer is only sold say for 5 months of the calendar year... e.g. August, September, October, November and December. Nominally the data is monthly. In fact the data is five period, i.e. 5 readings per “sales season”. It looks to me that you have 5 months (major period = 5) of the year of daily data.
I would fill in (impute via daily effects) the missing values within the 5-month time interval and use a model that captured daily effects and major period effects.
If you wished to post that data set, I would be glad to help further.
