In the BCa Bootstrap Method for Confidence Intervals, we have those two indices (instead of the 2.5 and 97.5 ones whose choice is the one in the Percentile Method): $$ k_1 = \Phi \left( \hat{z}_0 + \frac{\hat{z}_0 - z_{0.025}}{1 - \hat{a}(\hat{z}_0 - z_{0.025})} \right) \times B $$ and $$ k_2 = \Phi \left( \hat{z}_0 + \frac{\hat{z}_0 - z_{0.975}}{1 - \hat{a}(\hat{z}_0 - z_{0.975})} \right) \times B $$ where $z_{0.025}$ can be found in any complementary cumulative $Z$ table.
But what if those indices aren't integers? Does that happen? I couldn't find any reference on this.
Should I just round those?