SimuMatic

[vnigam]

I am trying to use the SimuMatic Module of Weibull++ version 7. I was going through the illustrated example section 18.4 in the User Manual. The solution says that there is the least deviation of the true parameter line when the parameters were estimated using RRX method. What is this differential? Is this the differential of the true parameter line and the median line or something else? Could you explain please? In either case, how does this tell me that a particular estimating method is a better one?

Thanks

VN

[David]

Hi vnigam,

Basically, you are just graphically comparing the median line to the true parameter line. Through simulation, the median line is closer to the true parameter line for RRX. Therefore, RRX is determined to be the more appropriate analysis method.

I hope this helps.

[vnigam]

THanks David.

But why the median line? Why not the average parameter line? Could you explain the significance please?

VN

[David]

Rank regression on X (RRX) is an unbiased estimator for Beta of the Weibull distribution and sigma in Normal and Lognormal distributions. But, for maximum likelihood estimation (MLE) these parameters are biased. Unbiased means the expectation of the estimated parameters are the true parameter values. So for a larger sample size, the expected value is the mean line and the mean line is very similiar to the median line. However, if the sample size is small, the mean line can be far from the true line even if there are a small number of outliers. Therefore, using the median line is better for both small and large sample sizes.

I hope this helps.

[vnigam]

Yes it does. Thanks

VN