Appendix

A part of QCP of student 1 in Statistical Simulation course (bold items are my response to the student):

To tract the distribution of statistic, such as sample mean is distributed as normal with mean mu and variance sigma^2/n. This may not be the best example. Usually, you know this by theory (e.g., by CLT). In these cases, you dont need simulation.
Parameter estimation and comparison: In order to decide which estimator is better, mean square error (or variance, if estimator is unbiased) is calculated.

Note: In the previous comment paper, at the graph that I draw to investigate randomness, you asked what I was supposed to see: There should be no pattern to say that data is random. You suggested drawing lines. In the homework, I see that lines are useful to see the randomness. Good!

A part of QCP of student 2 in Statistical Simulation course:

Last lecture we have firstly started importance sampling. In this method we are overcoming the problems while estimating = E [g (x)]. These problems are having large variance of g(x) and / or difficulties in simulating random vector X. Here h(x), importance function, is employed. This function should be chosen in a way that simulation of r.v. from h(x) should be easy and also that would give smaller variance.

In jackknife method, are there any conditions about the sample observations like the ones in bootstrap such as independence or correlation between x variables? You have stated that bootstrap is a special case of jackknife method, so the same conditions for x observations are also valid for jackknife method? This should be correct. My intuition is that jackknife will fail in similar situations when bootstrap fails. The only written material I have noticed about when jackknife might fail and bootstrap will work better is when the statistic is not smooth, i.e. when small changes in data will produce large changes in the statistic (Martinez, W.L. and Martinez A.R. (2002). Computational Statistics Handbook with MATLAB, pg. 245-247.)