I will be able to provide a clearer answer but I am not sure what you mean by "subsample of them according to the normal distribution".
1) If you use randn(1,100) it will generate normal random variables from a standard normal distribution (0 mean, 1 sigma), but this has nothing to do with your data.
2) For your data you will have to fit a distribution and then generate random samples from that distribution.
Depends on what you want, please explain your use-case and what you want to do.
