On Saturday, November 16, 2013 11:48:03 PM UTC, Rich Ulrich wrote: > On Sat, 16 Nov 2013 14:34:56 -0800 (PST), Steve > > > > > > >I want to program a random distribution in the range 0.0 to 1.0, similar to a normal distribution, but with the centre point shifted to, say, 0.25. > > >The quality of the randomness isn't crucial, as long as it is shifted. > > >Given that I can create a random normal-ish distribution, which is good enough for my purposes, by averaging a few regular Random(0,1), is there a technique to make a shifted version? > > > > > > > Problem number 1: "similar to normal" is not at all > > specific enough, especially when you want to assign > > a small range. Normal has infinite extent to both > > sides. > > > > If you just want a standard normal with a new location > > and variance, you could start with standard normal, > > mean=0.0 and variance= 1. > > > > Then multiple by a number to limit the range; and > > add a constant to adjust the mean. > > > > If you wanted to confine a normal to [0,1] while > > preserving the first 3 SD of range, you would take > > a normal deviate and divide by 9, then add 0.5. > > Trim or truncate (whichever you arbitrarily decide) > > whatever draws exceed the range. > > > > If you want a skewed distribution between 0 and1, > > you might look at the beta. > > > > -- > > Rich Ulrich
Thanks Rich. I don't understand 100% at the moment but there's enough there now that at least I know what to search for, study, and try. Specifically, I don't know what 'normal deviate' means. If it means the rate of change of the normal curve, I don't know how to calculate that.