10.06.2016 Views

eldo_user

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Eldo Control Language<br />

Statistical Processing<br />

_simu_get_mcsens<br />

Task function category: Statistical Processing Functions<br />

Returns the sensitivity list of the elements of the array array for a given sampling.<br />

Usage<br />

_simu_get_mcsens(array, "GLOB"|"GLOB_ACCURATE"|"LSS" [,<br />

matrix_raw_sampling])<br />

Arguments<br />

• array<br />

The array of values (typically the values of a given extract for each run) to compute the<br />

sensitivity.<br />

• "GLOB" | "GLOB_ACCURATE" | "LSS"<br />

Specifies the sensitivity methodology:<br />

"GLOB"<br />

Eldo computes global sensitivities extracted from the results of the random sampling.<br />

This is faster than "GLOB_ACCURATE" but results may be less accurate.<br />

"GLOB_ACCURATE"<br />

Eldo computes global sensitivities extracted from the results of the random sampling.<br />

This is more accurate than "GLOB" and can have a large CPU cost depending on the<br />

circuit. The CPU cost is related to the number of sensitivities, number of Monte Carlo<br />

runs, number of parameters with Monte Carlo variation, and number of extracts.<br />

"LSS"<br />

Specifies the large scale screening algorithm to identify the principal parameters<br />

during a Monte Carlo analysis. This is useful when the number of variables is much<br />

larger than the number of runs allowed. A class of Least Squares solvers based on the<br />

principle of homotopy methods is used.<br />

• matrix_raw_sampling<br />

The matrix storing all the raw sampling values computed during the Monte Carlo<br />

simulation. If n runs were performed and there are p statistical parameters, then<br />

matrix_raw_sampling[i;j] must be the raw sampling value of the j-th parameter for the i-th<br />

run. The <strong>user</strong> can call _simu_get_raw_sampling() to easily get this matrix (this is what is<br />

done by default). Be careful not to use _simu_get_sampling() otherwise sensitivities may be<br />

wrong.<br />

Description<br />

The result is a matrix: the first line contains the index of the important parameters (the first<br />

being the most important parameter). The second line contains their importance percentage (1.0<br />

for 100%). Will return an empty array if not enough Monte Carlo simulations have been run.<br />

Eldo® User's Manual, 15.3 1065

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

Saved successfully!

Ooh no, something went wrong!