My question is whether it is appropriate to use Independent T Tests on distributions generated through bootstrapping.

To be more specific: I used bootstrapping to generate 1000 subsamples for each of two observed distributions from different samples. I then used an SPSS macro to compute a diversity measure for each subsample (HREL).

For the bootstrap distribution of these diversity measures, for sample 1 I found a mean of 0.854 with a Standard Deviation of 0.044. (NB The maximum diversity score is 1.)

For sample 2 I found a mean of 0.771 with a SD of 0.052.

Can I use an Independent T Test to test whether this difference in the Mean is statistically significant? (I’ve run this and the results appear to suggest that they are, even though the 95% Confidence Intervals for both distributions overlap [i.e. CI: sample 1 – 0.768 to 0.940; CI: Sample 2 – 0.673 to 0.869] ).

I would welcome any advice on whether this is an appropriate thing to do and whether a difference in the mean can be deemed statistically significant, even when confidence intervals overlap.

Thanks in advance.