Dealing with definition changes mid time series


New Member
Can anyone point me in the direction of how I can deal with an abrupt change midway through a time series. The change is suspected to be due to data collection changes, not genuine and it is a significant shift in level but not trend. by 'deal with' I mean I need to be able to assess how things have improved since the start of the time series, even though we think the counting changes halfway through.


Well-Known Member
Hi Klik,

You probably have two options.
If you know the way the counting was changed, you may adjust the old data count to the new count.
Otherwise you may analyze the old data or the new data, or both separately


Fortran must die
Well first you want to analyze if there is a structural break at that period. Something like a chow test.

If there is one you normally analyze the data separately within the different periods. Essentially you have a new process so past data is of little value before the break.