Event prediction

[feed...@laathetmeweten.nl]

Hi group ... hopefully this is the right place to ask this question.
If not ... be gentle ...

Over the last two years I've collected a huge amount of information on
when certain events occur.

For example:

An event has occurred once in week 1, twice in week 4, once in week 7
and twice in week 10. It is likely that the event will occur again in
week 13, and probably only once, but what statistical method should I
use to predict this?

Unfortunately my knowledge of the statistical field is limited, but as
a software engineer I'm not afraid to delve into this field ... if
only I knew where to start. Hopfully you can give me some magic words
so I can start my search.

- daniel

[J. Horikx]

feed...@laathetmeweten.nl reageerde als volgt:

Because you are a software engineer, the best thing to do is to
write a simulation program. When you have 12 weeks and 6
occurences, then the probability per hour of the thing to occur is
1/336. Now you can simulate the weeks and occurences. To simu-
late one week you make a loop of 168 (7 days, 24 hours pro day)
hours and let something occor with a probability of 1/336 each hour.

(When you want the technical term, search "Poisson distribution",
but I think my approach is more intuitive and easier for you)

Of course you have to simulate many weeks to see a distribution
where you can be confident about.

JH

[J. Horikx]

J. Horikx <jVERWIJDE...@chello.nl> reageerde als volgt:

>(When you want the technical term, search "Poisson distribution",
>but I think my approach is more intuitive and easier for you)

I forgot, binomial distribution is also possible in this case, and
is even better (because its easier to understand and implement)

JH

[feed...@laathetmeweten.nl]

On Oct 7, 12:49 am, J. Horikx <jVERWIJDERDIThor...@chello.nl> wrote:
> J. Horikx <jVERWIJDERDIThor...@chello.nl> reageerde als volgt:

Thanks JH! Poisson rings a bell and I'll have a look at binomial
distribution.