Question: Sample from a distribution fitted to some data

I have a small data sample of some occurrence in time.  For an example lets take the time it takes kids to finish their lunch to determine wether a school has given them enough time for recess.

The times are (20,15,12,18,15,13,13,10,14,14,15,17,8,22,16) in minutes.

What distribution should I use? 

Should I just calculate the mean and std. dev. from the data and generate a normal distribution, like below?
with(Statistics):
a:=[20,15,12,18,15,13,13,10,14,14,15,17,8,22,16]:
m:=Mean(a):
d:=StandardDeviation(a):
X:=RandomVariable(Normal(m,d)):

Sample(X,3) # generate a few samples

I could also randomly choose from the observed 15 data points. 

combinat[randcomb](a,3) # randomly pick 3 of data points

Which method would be better?  Is randcomb following any kind of distribution?