Question around inspection event

dyadav

Note: You should only use events as predictors that happened before the remote and on-site inspections for the particular year have started (for example, the 27th of June 2015). 

We see that the minimum timestamp of inspection for 2015 is June,28 across all the cases for 2015 which is exactly one day after the stated date in the above note. 

Does the above statement mean that we can use all the events occurring before the minimum timestamp of inspection across all the cases for a particular year? Or does it mean something else?

In case our understanding is incorrect, can we get the date for the inspection in 2016 and 2017 before which we can use the events across all the cases for these years respectively.

fborchert

Your understanding is absolutely correct :-)

dyadav

If events occurring before the minimum timestamp of inspection across all cases can be used for a particular year, then this minimum timestamp is Feb,16 for 2016 and Feb,07 for 2017.
And we see that only 1 or 2 events occur before these dates in 2016 and 2017 respectively.

Should we use these minimum timestamps of inspection for filtering the events that could be used as predictors for Question 2? Or any other date (for 2016  as well as 2015) can be used?

fborchert

That's a very interesting observation.

We obviously have some outliers in the data where inspection documents are created much earlier in the year.

If you ignore these outliers, reasonable dates appear to be: June, 2nd for 2016 and May, 31st in 2017.

It would also be interesting to see if these outliers are special from other perspectives, as someone obviously decided explicitely that these guys should be subject to inspection.