I would like to decompose the log of a time series. There will be time slots that are zero. In order to handle this I insert 'NA' for all of the zeros in the time series. Having a zero value is a very legitamate value. With those NAs in the time series stil requires an na.action argument because the default is na.fail which I don't want. I am afraid if I use na.omit it will skew the seasonality and trend because what if there are seasonal components which are zero? Anyway I was looking for recommendations for what the value of na.action should be for stl. Ideally I would like to just keep the NAs there, not flag an error yet still have those values "count" for the seasonal or trend calculations. Thank you. Kevin