Thread Subject: random data

hi,
i've got such a question: i've got an ellipse (know
minor/major axis and it's centre).

how can i generate random data that lies all-around my
ellipse (but not on it!)? ok, some
points can lie on it but not all of them.

      kind regards

In article <g408tr$7cb$1@fred.mathworks.com>,
mark b. <john.doe.nospam@mathworks.com> wrote:

>i've got such a question: i've got an ellipse (know
>minor/major axis and it's centre).

>how can i generate random data that lies all-around my
>ellipse (but not on it!)? ok, some
>points can lie on it but not all of them.

Should the data be inside or outside the ellipse? If outside, then
what kind of probability distribution do you want for the distance
away? e.g., probably "most" of the points should be near-by but
perhaps there is a 1 in a million chance that the random point
will be (say) 1000 units away? Single-tailed normal distribution
perhaps??
--
  "The slogans of an inadequate criticism peddle ideas to fashion"
                                              -- Walter Benjamin

the best option will be if some of these points lie inside
the ellipse and some of them outside.
about probability distribution i don't care that much.
i mean, they should lie quite close to my ellipse, but it
will be nice if at least few points lies further. but 1000
units away is too far :) for example, let's say max 7 units
away. do You have any idea how to do this? does matlab has
any function to generate it in some simple way?

 kind regards

"mark b." <john.doe.nospam@mathworks.com> schrieb im Newsbeitrag
news:g409pu$hp1$1@fred.mathworks.com...
> the best option will be if some of these points lie inside
> the ellipse and some of them outside.
> about probability distribution i don't care that much.
> i mean, they should lie quite close to my ellipse, but it
> will be nice if at least few points lies further. but 1000
> units away is too far :) for example, let's say max 7 units
> away. do You have any idea how to do this? does matlab has
> any function to generate it in some simple way?
>
> kind regards
>

If you are not caring too much about the probability, the very simplest way
I could think of would be to calculate points randomly on the ellipse and
add some noise to it:

n = 100;
t = rand(n,1)*(2*pi);
x = 2*cos(t)+0.07*randn(n,1);
y = 3*sin(t)+0.07*randn(n,1);
plot(x,y,'r*', 2*cos(linspace(0,2*pi)), 3*sin(linspace(0,2*pi)),'b-')

Titus

thx:)