3
$\begingroup$

When people speak of "simulation variance", what does this mean? Does simulation variance disappear as $N \rightarrow \infty $?

Are models that have less simulation variance for a fixed $N$ considered more "robust"?

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ Could you give references? The only concept I can think of is the error variance, which clearly does not go away when N increases. $\endgroup$
    – mpiktas
    Oct 24, 2011 at 12:03

2 Answers 2

Reset to default
2
$\begingroup$

Simulation variance is brought about by the fact that estimating expectations (integrals that are part of likelihood functions) by averaging a bunch of realizations of the respective random variables is not precise for a finite bunch. It gets smaller with the number of simulations and not related to the sample size, N.

Economic models may not have less simulation variance, they assume integrals to be computed exactly, so it is a separate issue.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Thank you so much. I discovered a half-answer to this already, but what you said clears things up a lot. Simulation variance is due to the inexact calculation of integrals. well said! thanks again. $\endgroup$
    – d_a_c321
    Nov 17, 2011 at 18:04
1
$\begingroup$

My answer is like Alex's. Simulation variance appears in all analyses that apply Monte Carlo. there is nothing special about structural economics. The simulation results are based on random sample of size N being drawn from a population distribution. The distribution this sample (called the empirical distribution) is different from the population distribution because N is finite. If you took you did it again for another sample of size N you would not get the exact same empirical distribution and hence the simulation result would be different. This difference from one sample to the next has its variability characterized by the simulation variance. Now it is well known that for indpenedent identically distributed random variables as N increases the empirical distribution converges to the population distribution. So increasing N leads to distributions that are close together and hence the simulation variance is getting smaller and going to 0 and N approaches infinity.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.