Calculate bootstrap sd, disregarding some bootstrap estimates (when model is overfitted)

Question

I want to calculate bootstrapped sd disregarding some of the values in t* (when e.x. bias=0, which happens when the model is overfitted).

I get the standard deviation