10.06.2016 Views

eldo_user

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Monte Carlo Analysis<br />

Monte Carlo Basic Features<br />

may be approximated with the summation:<br />

This property implies that a convergence test can be constructed and that confidence bounds on<br />

the computations can be provided. In Eldo it is possible to use the .MC AUTOSTOP parameter<br />

to control the run length for estimating the average value E(H(X)) (this can save CPU time).<br />

For example, a convergence test based on confidence interval techniques can be defined with<br />

the extract function MCCONV as follows:<br />

* EXTRACT SECTION<br />

* THE PERFORMANCE CRITERIA OF INTEREST IS:<br />

* THE OSCILLATION FREQUENCY: OFREQ<br />

.EXTRACT TRAN LABEL=OFREQ 1/(XDOWN(V(1),2.5,END)-XDOWN(V(1),2.5,START))<br />

* MONTECARLO ANALYSIS<br />

.MC 1000 AUTOSTOP=’mc_conv_freq’<br />

.EXTRACT MC LABEL=mc_conv_freq MCCONV(OFREQ, AVG, CONFIDENCE, 100, 0.95)<br />

where the expression MCCONV(OFREQ, AVG, CONFIDENCE, 100, 0.95) is a Boolean<br />

value, which is true if some specific criterion is satisfied for the sample mean value AVG.<br />

Example with multiple MCCONV extracts:<br />

* MONTE CARLO ANALYSIS<br />

.MC 1000 AUTOSTOP='mc_conv_avg && mc_conv_std'<br />

.EXTRACT MC LABEL=mc_conv_avg<br />

+ MCCONV(V1,AVG,CONFIDENCE, 100, 0.99, 0.0004, 0)<br />

.EXTRACT MC LABEL=mc_conv_std<br />

+ MCCONV(V1,STD,CONFIDENCE, 100, 0.99, 0.0003, 0)<br />

Note that the Central Limit Theorem says that the sample mean converges to the expected value<br />

at the rate O(n −1/2 ) which means that in practice an extra digit of accuracy requires 100 times as<br />

many simulations. The convergence is therefore slow compared to more efficient numerical<br />

methods such as Riemann quadrature or Simpson method, but the convergence of the MC<br />

method is completely independent of the dimension of the input space.<br />

Related Topics<br />

Specifying a Monte Carlo Analysis<br />

CONFIDENCE Technique<br />

418<br />

Eldo® User's Manual, 15.3

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!