suppose there is a probability x that an event happens (and clearly 1-x it doesn't). The catch is, x is not constant, it depends on how many times x has happened before. In other words, with a probability x, the event happens and x increases. With a probability 1-x, the event doesn't happen, and x remains the same. Hence the probability of the event happening increases in how many times the event has happened.
I know this isn't a lot of detail. I am just looking for a suggestion of how to begin modelling something like this. I don't know where to even start. Does anybody have any quick comments about where I should look?
Thank you very much,
May 19th 2010, 07:17 AM
This is called a reinforced process. Simplest example is Polya's urn: Consider an urn with balls of two different colors. Each time we pick a ball, we put it back together with a new ball of the same color (hence this colors becomes more likely).
What you're asking for is quite vague. For any sequence of numbers in , you can define a process with values in such that, given , the probability of is where . If is increasing, this is a general reinforced process. The choice of this sequence is the only data you need.
May 21st 2010, 05:36 AM
Thanks for your response,
Basically every time the event occurs, the probability of it happening again increases by x%. For example, suppose the probability of drawing a red ball is .5, then if you draw a red ball today the probability you draw a red ball tomorrow is .55 (if x=.1) etc... I want to express as a function of time t, the expected number of times the event has happened at any given time. Is this possible to do? Thanks in advance,