Re: populate a list with random numbers from normaldistribution?
- To: mathgroup at smc.vnet.net
- Subject: [mg49931] Re: populate a list with random numbers from normaldistribution?
- From: koopman at sfu.ca (Ray Koopman)
- Date: Fri, 6 Aug 2004 03:09:56 -0400 (EDT)
- References: <cetek6$6fe$1@smc.vnet.net>
- Sender: owner-wri-mathgroup at wolfram.com
sean kim <sean_incali at yahoo.com> wrote in message
news:<cetek6$6fe$1 at smc.vnet.net>...
>
> What I wanted to do is to pick sets of random Integers
> in the ranges {10000, 99999} from the Normal
> distribution such that, the mean is 50000 and the
> standard deviation is 25000.
>
> or is that not normal distribution? will that be a
> custom distribution?
>
> if that isn't possible, then I want to pick set of
> random intgers ranging from {10000, 99999} such that
> the variance and the mean and standard deviation do
> not change across different runs.
>
> I think i'm supposed to use SeedRandom here but this
> caused me problems also.
>
> so I need to design a routine that will pick random
> integers in a given range with a given distribution
> with the SAME mean, std dev/ and variances across
> different runs or simulations.
You want integers, but normals are real. However, this
is not serious, because the quantization interval (1)
is so small relative to the standard deviation (25000).
The center of the interval (55000) is .2*sd away from the mean. Do
you really want that much asymmetry, or would a mean of 55000 be ok?
With the mean at 55000, the interval limits are mean +/- 1.8*sd,
which captures only ~93% of the normal distribution. Do you really
want to cut off the tails of the distribution? It certainly won't
be very "normal".