17.01.2015 Views

GPS-X Technical Reference

GPS-X Technical Reference

GPS-X Technical Reference

SHOW MORE
SHOW LESS

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

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

Optimizer 386<br />

Portmanteau Sub-Section<br />

Portmanteau test on weighted residuals (ON - OFF): This switch turns the<br />

Portmanteau test on or off. The Portmanteau test is used to detect trends in the weighted<br />

residuals. The Portmanteau test is designed for data taken in sequence (e.g. time or<br />

space). If trends are present, the residuals are not independent. This violates one of the<br />

assumptions of the maximum likelihood method and indicates that the model does not<br />

account for all of the non-random variability in the data. An appropriate message is<br />

printed depending upon the outcome of the test. This test is only reported when the<br />

maximum likelihood objective function is used. It does not affect the outcome of an<br />

optimization run.<br />

Maximum number of lags used in portmanteau statistic: By default, the Portmanteau<br />

statistic involves autocorrelations at lags up to half the length of the time series. This<br />

setting can be used to impose further restriction on the number of autocorrelations taken<br />

into account. A large value effectively disables this option. This setting does not affect<br />

the outcome of an optimization run.<br />

OPTIMIZATION STRATEGIES<br />

Before using the optimizer tool to estimate optimal parameter values or optimize operating<br />

conditions, it is usually best to experiment with manual adjustment of the optimization variables.<br />

By conducting interactive simulations you can observe the effects of the model parameters on the<br />

response variables of interest. With this information you will be able to make better judgments on<br />

appropriate variables to use in a parameter estimation or optimization run.<br />

For example, you can set up an interactive simulation with slider controls for the parameters and<br />

then try adjusting these variables to try and achieve a visually acceptable fit. You can plot actual<br />

data along with the target variables so that you can compare simulation and actual data. This<br />

approach is useful for generating starting guesses for parameter estimation and process<br />

optimization runs.<br />

<strong>GPS</strong>-X <strong>Technical</strong> <strong>Reference</strong>

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

Saved successfully!

Ooh no, something went wrong!