How to use bootstrapping to compare data converted into %s above threshold for statistical significance I was looking at velocities of two different data sets ($a$ and $b$). The idea is to look at velocities greater than a certain threshold, $v_b$ and the find the percentage of moving objects. So I take the sum of all velocities ($v_{\rm total}$) and the sum of velocities greater than $v_b$ ($v_{b\ {\rm total}}$). The percentage of moving objects then are $v_{b\ {\rm total}}/v_{\rm total}$. So I bootstrap the percentages using this method. Now say data $a$ has a percentage mean of 58% and data $b$ has a percentage mean of 46%. How do I compare the statistical significance between them?
 A: GENERATE THE BOOTSTRAP SAMPLE.
For looking at the difference between two means you take the sample differences of the observations. Get the n differences.  Then generate a bootstrap sample by sampling with replacement from these n mean differences.
Next specify how many samples of size n you will generate. Typical you can pick 1000 but you cab take more. The bootstrap samples give you a histogram for the bootstrap distribution of mean differences. 
COMPUTE A BOOTSTRAP CONFIDENCE INTERVAL
Then pick say Efron's percentile method bootstrap confidence interval using a 95% confidence level. Your null hypothesis is probably that the mean difference is 0.  You will test at the 5% level by checking whether or not the 95% bootstrap confidence interval contains 0.
NOW USE THE CONFIDENCE INTERVAL TO TEST THE HYPOTHESIS
Reject the null hypothesis if the interval does not contain 0. You obtain the p-value by determining at what significance level the interval would just barely contain 0. 
KEEP THIS IN MIND
The bootstrap is a nonparametric method. So you don't need to assume the data is normal and the bootstrap histogram need not look normal.
