Illogical values in 95% CI

We have a 95% CI for a count variable that includes a negative value (the lower limit of the interval is negative). Counts can't be negative, so this isn't a logical value. We propose to change the negative value to 0. Is there a reference for justifying this? It seems very logical to do this, but we were asked for a reference to support the idea.


Ambassador to the humans
It sounds like you're relying on a method/model that is making assumptions that don't hold in your data. Can you describe how you're making your confidence intervals?


Not a robit
I second @Dason inquiry. This is also seen when people use normal CI calculations with percentage value data, which are bound within 0 and 1. Then people wonder why they get CI values outside the bounds. You need the right CI formula for your data.