# Patrick

less info
reputation
17
bio website location age member for 1 year, 6 months seen Oct 28 at 10:36 profile views 17

I'm a mathematical computer engineering student with various interests.

# 45 Actions

 Jul22 awarded Yearling Jun20 awarded Teacher Jun12 awarded Scholar Jun12 accepted Gauss-Markov theorem: BLUE and OLS Jun12 answered Gauss-Markov theorem: BLUE and OLS Jun3 comment “Monte Carlo Kalman Filter” vs Unscented Kalman Filter I don't know unscented KFs much, but the MCKF sounds a lot like the EnKF. You might want to check that out, and possible comparisons with the unscented KF. Apr25 asked Under what conditions do Bayesian and frequentist point estimators coincide? Apr25 comment What if the MVUE depends on the parameter? @whuber, I'm asking this question because I'm looking for conditions under which Bayesian (with a flat prior) and frequentist point estimators, derived as optimisers of expected losses, coincide. Do you happen to know any references on this? Apr25 revised What if the MVUE depends on the parameter? added 566 characters in body Apr25 comment What if the MVUE depends on the parameter? Yes, I agree. It cannot. However, its definition, without the uniform bit, does not guarantee this. (And I insist again that my definition is the same as the "conventional one".) Which implies that its existence, is not guaranteed. It seems weird then that a frequentist would completely have to abandon the MVUE criterion in these "degenerate" cases, while a Bayesian could easily make sense of them too by taking expectations over the parameter as well. Apr25 comment What if the MVUE depends on the parameter? It is the same as the definition on the wiki page you link to, except that the wiki page discusses uniformly MVUEs. I.e. UMVUEs are the MVUEs for any theta, which implies that their formula does not explicitly depend on theta. The article doesn't take the existence of UMVUEs for granted, however, which I think means that in some cases one does indeed obtain an MVUE that does depend on theta. So it seems that a frequentist would then be at an impasse, and would have to use some other criterion for deriving some optimal estimator. Apr25 asked What if the MVUE depends on the parameter? Apr25 comment Condition for Law of Large Numbers, Monte Carlo Yes. Correct. And the same thing goes for h as well as for f. Apr24 comment Condition for Law of Large Numbers, Monte Carlo Sorry, no. My sentence "sample, x, is drawn from g that is outside the support of h" does not indicate that I suggest that the support of h encompasses that of g. I'm pretty sure if you think about the answer and my second comment you will understand. Apr24 comment Condition for Law of Large Numbers, Monte Carlo Well, did you get it? @Delvesy Apr23 comment Getting a posterior belief from noisy observations Why exactly does this rule out factor analysis? Apr22 revised Gauss-Markov theorem: BLUE and OLS added 4 characters in body Apr22 revised Gauss-Markov theorem: BLUE and OLS More details Apr22 comment Under which assumptions does the ordinary least squares method give efficient and unbiased estimators? You might want to see my related question, and clearly the answer seems to be "yes", but only among linear estimators. Apr21 comment Queries on the Bayesian method Thanks for the insight of the usefulness of using different priors.