Max
Reputation
Next privilege 75 Rep.
Set bounties
 Apr21 comment Gorgeous diophantine equation I found this problem in the ancient unnamed academic tutorial about the number theory. And I can't let go... Apr21 asked Gorgeous diophantine equation Mar2 revised What problems are related with the following type of FDE with delay? deleted 77 characters in body; edited title Mar2 revised What problems are related with the following type of FDE with delay? added 266 characters in body Mar2 revised What problems are related with the following type of FDE with delay? added 266 characters in body Mar1 awarded Benefactor Mar1 accepted The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ Mar1 revised What problems are related with the following type of FDE with delay? Clarifying Feb26 revised What problems are related with the following type of FDE with delay? added 11 characters in body Feb26 asked What problems are related with the following type of FDE with delay? Feb25 revised The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ clarifying Feb25 comment The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ I'm a little confused by the variance formula. Actually, we don't know the true $\alpha$. Can we substitute $\alpha^*$ by the its estimated value? Feb24 comment The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ Alexander, it's very good extension, thanks. The 3rd estimator is little bit surprising and at first sight it is unbiased. Could you specify the variance for it? Feb22 comment The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ Alexander, what about the reliability of the suggested estimate? Is it unbiased? What is the variance? Feb22 revised The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ deleted 2 characters in body Feb22 awarded Promoter Feb22 comment The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ It's hard to believe thar PDF of $\epsilon_i(\hat \alpha)=(x_i \frac{\sum_i x_i/\delta_i}{\sum_i x_i^2/\delta_i} - 1)/\delta_i$ is most similar to $N(0,1)$. The goal of your solution is to minimize the variance of $\epsilon_i$ rather than to make it equal to 1. Furthermore, you don't take into account the mean of $\epsilon_i(\alpha)$ which should be close to zero. Feb22 comment The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ I think that a good estimation of $\alpha$ shoud make PDF of $\epsilon_i$ most similar to $N(0,1)$. Feb22 comment The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ The Gaussâ€“Markov theorem for the least squares estimator (LSE) assume that residuals are i.i.d. and we don't know the true form of its distribution. In that case will get the best linear unbiased estimator for a linear regression model which gives us the lowest variance of the estimate. In our case a distribution of residuals is well known. We can't use LSE because we don't need to minimize the variance of residuals. Feb21 comment The $\alpha$ estimation for the model $x_i = \xi_i \cdot \alpha$ Thanks, but why are you use $\sum_i \epsilon_i^2$ as the goal function? We already know the distribution of $\epsilon_i$. We shouldn't to minimize the sample covariance of $\epsilon_i$ because it is already known. Moreover, in your case the sample mean of $\epsilon_i$ will sufficiently far from zero.