The problem is that some hospitals miss submissions on some days (true) missing, whereas some hospitals submit a false zero for some of the data items on some of the days.
I want to create a national time-series adjusted for these missing data and false zeroes. Can anyone suggests if there is a recommended approach for this type of issue please?
At the moment I am replace zeros with missing using a semi-manual process that is far from ideal. I'm then inputting missing values using a polynomial if the hospital has more than 4 observations in total and using
Code:
poisson y c.day##c.day i.hospital
Many thanks for any insight you can offer.
Rob
0 Response to Detecting false zeros in panel data
Post a Comment