Time Series with fpp3

Machine Learning and Modeling
1

I am working on a time series analysis and am using the fpp3 package. I have a tsibble with 3 timeseries in it. I have checked that all 3 time series have the same 37 time period entries and there are no null or anything.

When I try to use the SEATS, STL, X11 decomposition it works perfect for 1 of the 3 series but for the other 2 it errors out.

Errors:

  • A model estimation error has occurred during AIC testing within the automatic model identification
    procedure. The error message appears below.
  • Regression matrix singular because of Leap Year. Check regression model or change automatic outlier
    options i.e. method to addone or types to identify AO only.

I can run a classic decomposition and when I pull the components from ETS models they are displayed and it is interesting because the dates don't line up across all the components. The data series and seasonal goes monthly from Jan 2017 to Jan 2020 however the trend and random components only go from July 2017 to July 2019.

Referred here by Forecasting: Principles and Practice, by Rob J Hyndman and George Athanasopoulos

2

Hello,

First, a reprex, or data would be helpful. Second, there are functions in fpp3 package that check for time series data structure requirements. For example, has_gaps() checks to see if you have gaps in your time series. There is are several other similar functions. I would start there.

3

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.