Have derived estimators that are more efficient than the classical ones.I Run Monte-carlo simulations to study the performance of the estimators. Do i need a real data to test the new estimators?
Deriving an estimator with better properties is one thing (however, I think that the better performance should be proved mathematically, not just shown by example in a simulation); demonstrating that it is worth for real problems (where assumptions hold only approximately and models are not (can not be) specified correctly) is a different thing.
If you want to demonstrate that your estimator really works better, you have to show a real application where your estimators are superior. And you will need a panel of different data sets to demonstrate that the superiority is not attributable to a special feature in one particular set of data.
Deriving an estimator with better properties is one thing (however, I think that the better performance should be proved mathematically, not just shown by example in a simulation); demonstrating that it is worth for real problems (where assumptions hold only approximately and models are not (can not be) specified correctly) is a different thing.
If you want to demonstrate that your estimator really works better, you have to show a real application where your estimators are superior. And you will need a panel of different data sets to demonstrate that the superiority is not attributable to a special feature in one particular set of data.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Probability