When you have a large set of data, it can be resource consuming to calculate standard deviation at every step. So why not calculate a recursive std deviation at step n given:

• mean at step n
• value of added point at step n
• mean and std at step (n-1)

PS1: We'll always consider a vector of length 2 to which we add a third value

PS2: There are two common definitions of stdandard deviation. If you look at the help section of matlab, i used the second definition