I'm presently working on a paper in which bootstrapping is to be used. We plan to bootstrap our model's predictions and present these, as well as confidence intervals for the model predictions.
When I plot the mean of the bootstrap replicates for my predictions, they are off centre for some intervals. Take a look below: The prediction (black line) is off centre from the 50% confidence interval. Now, I know that these intervals should not be symmetric due to the log link, however this also implies that the upper limit should be farther from the prediction than the lower limit. This is ameliorated when I instead plot the original model predictions (red).
Do I report the mean of the bootstrap replicates for my model, or, do I report the original model predictions + bias corrected bootstrap confidence intervals?