Although the **dround** "distribution" seems conceptually simple, I get an initialization error message that makes me wonder if I'm using it right. I'll illustrate with a minimal example: given an observed integer **x** we want to infer the real value **truescore** which is know to be in the interval 0-100.

truescore ~ dunif(0, 100) x ~ dround(truescore, 0)

This works for x=50 (or more generally for x in the middle of dunif) but not for any other values (e.g. 51, 2, 65 etc.). It says *Error in node x, Observed node inconsistent with unobserved parents at initialization.* It seems strange that initialization values is necessary in such a simple case. Am I doing it wrong?

Furthermore, for x=50, the posterior on "truescore" behaves as if the observation was rounded to nearest integer for digits parameter below 2. (E.g. for 1, 0, -1, -100 etc. the posterior is x +/- 0.5). I would expect the posterior to broaden as the measurement resolution decreases and for digits=1 to give an estimate of +/- 0.05 rather than +/- 0.5.

Best,

Jonas