Note that in the scotgov_management dataset, I've converted all *'s (which denote <5 ?) to zero!!

[soniamitchell]

No description provided.

[IainM-BioSS]

Thanks for the flag on this, Sonia. Are we confident that all recorded * are indeed <5 events? How do they record missing data? Do they denote observations in the range 0<=x<5 or 0<x<5 .. ie are there also 'genuine' zeros? Some of the inference models could, in principle cope with censored data. What matters is that we distibguish, if possible, between zeros, <5 and missing values.

[soniamitchell]

I am pretty confident – but this will need to be checked. I could reinstate the * to distinguish between those values, 0s and NAs?

[IainM-BioSS]

No, it's fine for now, especially since you have flagged it as an important issue that users need to be aware of. But it is exactly the type of situation where we just need to be a wee bit careful. There is often an issue distinguishing between zeros and missign values, and the censored element just makes it more fiddly. If we are confident that there are no zeros.. ie that they are all included in the * terms, then coding them all as zero is very much the right thing to have done. But sense checking the data for zeros and missing values would be a good idea.

[soniamitchell]

I'll add an issue to the database