I am conducting a research for modelling risks in construction projects using Bayesian networks. The whole network is constructed by expert knowledge. In the process of introducing CPTs, how many expert panels are required or what is the sample size for this particular stage?
There is no simple answer for that. I think the most important is that the spread of data should be al low as possible. But if if you have two experts with total different perception, I think you can build to BBN to show that.
Thanks for your contribution,
The method undertaken for defining prior probabilities is based on five point likert scale through questionnaire. According to the scope of this research, the sample size is around 350 respondents and this is only for prior probabilities. Defining prior probabilities and CPTs, are not they in the same importance?
One researcher took 8 experts, another 20 and I found someone took 45 experts in CPTs calculation, but there was no justification. How confidential their assessment could be if there is no reference for this. is it that simple just taking tow expert in the field?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size