Avoid negative yield stress values about gooseepm

[tdegeus]

Sounds programmatically fair. Technically this could always happen even if the standard deviation is small. I wonder, however, if the absolute value is the right thing to do. We could also truncate at zero?

[laurentbugnard]

You are right, I suggest we use a truncated normal distribution. It would be the most neutral choice, as our friend wikipedia says: "The truncated normal is the maximum entropy probability distribution for a fixed mean and variance, with the random variate X constrained to be in the interval [a,b]." Since prrng doesn't have this feature yet, we could implement it. Or we could find a workaround: intuitively, I would say that simply refusing negative values and resample them would give the same result.