Specifying safety rules on holding level without final dummy argument gives unexpected result

[lverweijen]

Please specify
SDC tool used: tau-argus Version used: 4.2.0 Operating system used: Windows

---

I have been playing around with safety rules on an individual and holding level and I noticed an inconsistency that might result in a hard to catch error.

Given a microdataset with a holding supplied in the metadata.

My impression is that the following safety rules all work on the individual level although they should work on the holding level:

<SAFETYRULE> P(0, 0)|P(0, 0)|P(20, 1)  # No individual level supplied, because P(0, 0) are dummies.
<SAFETYRULE> P(20, 1)|P(0, 0)|P(20, 1)  # Rule repeated on individual and holding level
<SAFETYRULE> P(20, 100, 1)|P(0, 100, 1)|P(20, 100, 1)  # P/q rule with q=100%

The following safety rule does seem to work on the holding level:

<SAFETYRULE> P(20, 100, 1)|P(0, 100, 1)|P(20, 100, 1)|P(0, 100, 1)  # Extra dummy at the end

I have some recommendations:

• Not having a dummy after the holding safety rule shouldn't matter and the first three rules should work on the holding-level.
• Instead of working with dummy-arguments, I would recommend having separate tags SAFETYRULE and SAFETYRULEHOLDING to make the distinction. This is more obvious and makes it easier to write scripts around tau-argus.
• The form with 3 arguments isn't documented in the manual yet. It seems that the extra parameter of q=100% denotes the prior/posterior rule.

[lverweijen]

After testing some more, it seems to work anyway with the first 3 rules. It was probably just me that got confused somehow. I'll close for now.

[ppdewolf]

Note: the 3 parameter version is a left over. In the GUI it is no longer possible to use the p/q rule. The p/q rule is in some sense equivalent to a p%-rule for certain p, q and p.