I am an engineer without a strong statistics background. I'd like to do a Monte Carlo analysis to model uncertainty. I'm combining samples of measured data (N=~50) where I do not know each population distribution. I've done some basic things in R (looking at normality with Shapiro-Wilk, using the "fitdist" function with various distributions), but after some reading about bootstrapping and such, it seems more appropriate to just use the empirical distribution instead of a t-distribution, for example.
I imagine this is a loaded question, but when should I consider using a "defined" density function like a t-distribution instead of just using the empirical distribution?
Within R, how would I use the empirical distribution? It seems like this is very much bootstrapping (I've only recently learned about the term), so would I just be making use of the "boot" package? Or are there other ways to get to and make use of the empirical distribution?
Thanks. Any guidance or suggestion of reading materials is highly appreciated