Merging Predicted and Field data

[Merlino]

Why nobody from Reliasoft (or any other organization) has answer my question on 12-3-04 about this subject? Is it because it is very complex?
I was hoping that Mr . Pantelis was going to give me some advise.

[Pantelis]

Merlino wrt to my self I usually respond to postings that A) I have a good answer for, B) the question is clear enough so I can give you a good answer. Now with respect to your question, it is posed in two parts. The first part is a PRISM question – and to this am I not familiar with the formula. The second part talks about using a Bayesian approach. Here I am not sure I understand what you are trying to do. If your question is how do you check that a given distribution is applicable to a set of data -- then there are multiple statistical techniques that one can use (if you have data), but I am sure your familiar with these, so I a do not think that is what you are asking. Hope this helps

[Tarik]

If you running into problems when applying the analytical methods that you obtained from Dr Keccicoglu's books (which I haven't reviewed), you could consider a number of simulation methods such as the Markov Chain Monte Carlo, Gibbs and Sampling-Resampling methods.

[Merlino]

My question is simple: How do you merge predicted data and field data? Since predicted data is performed as a feasibility study for a potential project and field data is performed after the project is operational, and since predicted data is almost always more optimstic, is ther a way to merge the two?

[Rui Assis]

Suppose you have the following prior distribution (predicted data): (years-probability) 1999-0,1; 2000-0,2; 2001-0,4; 2002-0,2; 2003-0,1 and the following updating distribution (field data): 1999-0,01; 2000-0,03; 2001-0,05; 2002-0,07; 2003-0,09. Then you compute joint probabilities: 0,1x0,01=0,001; 0,2x0,03=0,006; 0,4x0,05=0,02;0,2x0,07=0,014; 0,1x0,09=0,009, whose sum is 0,05. Finally, you compute the posterior distribution: 0,001/0,05=0,02; 0,006/0,05=0,12; 0,02/0,05=0,4; 0,014/0,05=0,28; 0,009/0,05=0,18.
Hope this helps.

[Rui Assis]

In my previous message, please consider future years (2005; 2006; 2007; 2008 and 2009) instead of 1999 and so on. I gave an old example and forgot the necessary update. I am sorry for the mistake.