Small omega estimates and the covariance step

Dear all When NONMEM has estimated a very small value for an element in omega, the covariance step is omitted, with a message that the parameter estimates are near the boundary: PARAMETER ESTIMATE IS NEAR ITS BOUNDARY THIS MUST BE ADDRESSED BEFORE THE COVARIANCE STEP CAN BE IMPLEMENTED I have only put lower boundaries of 0 on theta, and none of the parameter estimates are close to 0, and I used $OMEGA 0.5 for the initial random effects when specifying an exponential variability. Apparently, in NONMEM 6 we have a more elegant error trap for when the covariance step is likely to be aborted. Great! The questions that I have are: - Since I have not specified a boundary for omega, does anyone know what the boundary is? Is it dependent on the initial estimate, or is it a fixed value? - Can I bypass this message and get the covariance step to be done - even if it subsequently aborts? - On a more philosophical note: One could imagine implementing an additive variability with very small values that could still be estimated precisely, eg when I am expressing concentrations in kg/mL. Would the implementation of this error trap mean mean that I have to multiply my observations with a high enough value to circumvent this? (and what would be 'high enough'?) Suggestions are much appreciated! Kind regards Peter