Related Topics: | ||
The Calculations page of Weibull/ALTA's Application Setup allows you to set the numerical precision used in the calculated results and plots, and whether to sort a data sheet before or after calculations.
Precision to be Displayed on Calculations
Use Displayed math precision to set the number of decimal places shown in results and plots. Please note that the tolerance on the precision is limited by Windows to 15 digits.
Use Scientific notation tolerance (10^?) to set the point at which the software will switch to scientific notation. For example, if the scientific notation tolerance is set to 5, then any number that is larger than 105 will be displayed using scientific notation.
Special Options on Location Parameter
Use this area to set the options regarding the location parameter (Gamma).
If you want to be notified if the location parameter contains a negative value, select Warn if location parameter is negative.
If you select Discard if location parameter is negative the software behaves as follows:
If you are using RRX or RRY to calculate the parameters:
For the 2-parameter exponential distribution, the negative location parameter is discarded and the data set is recalculated using the 1-parameter exponential distribution. The location parameter is displayed as zero.
For the 3-parameter Weibull distribution, the negative location parameter is discarded and the data set is recalculated using the 2-parameter Weibull distribution. The location parameter is displayed as zero.
If you are using MLE to calculate the parameters:
For the 2-parameter exponential distribution, the negative location parameter is discarded and the data set is recalculated using the 2-parameter exponential distribution with the location parameter constrained between zero and the first failure time.
If you have not selected the Use true 3-P MLE on Weibull option on this page of the Application Setup, the negative location parameter is discarded and the data set is recalculated using the 2-parameter Weibull distribution. The location parameter is displayed as zero.
If you have selected the Use true 3-P MLE on Weibull option, the negative location parameter is discarded and the data set is recalculated using the 3-parameter Weibull distribution with the location parameter constrained between zero and the first time-to-failure.
Select Reset if location parameter > T1 on Exponential to reset the value of the location parameter equal to the first time-to-failure if the location parameter is greater than the first failure time. The parameters will then be recalculated based on the new value of the location parameter. This option pertains to the exponential distribution for both rank regression on X (RRX) and rank regression on Y (RRY).
Select Use true 3-P MLE on Weibull to use a true 3-parameter MLE solution. The 3-parameter solution is inherently problematic for some Beta values and can fail to converge for some data sets. When not selected, Weibull++ uses non-linear regression to compute Gamma, and then computes the MLE of Beta and Eta using a transformed T' = (T - γ).
Select Use Theta parameter scheme on Weibull to replace the Weibull scale parameter Eta in all results with Theta, where Theta = Eta + Gamma.
Select Use extended Generalized Gamma to allow the Lambda parameter of the generalized gamma distribution to take negative values.
Other Options
By default, Lambda is used as the scale parameter of the exponential distribution. Select Use mean time for the exponential distribution if you prefer to use MTTF (1/Lambda) instead.
Select Sort before calculation to sort the times in ascending order before calculating the data sheet. Note that this will not affect the results of the calculation.
Use unbiased Std on Normal data
Use this check box to correct the MLE standard deviation (sigma) value for biasedness. Note that this setting applies only to the normal distribution with complete data.
Use special sort (place F before S)
The software always sorts entries based on failure times. In special cases where two or more failure times are identical to suspension times, the sort algorithm does not distinguish between them and sorts them based on the way they were entered. For example, if given a suspension time of 100 hours and a failure time of 100 hours, the results will vary depending on the order in which the data were entered and whether you are using rank regression. When the Use special sort (place F before S) option is selected and two identical times are encountered, the software will always put failures before suspensions. This involves a double sort routine and might be slightly slower than the standard sort.
Use plotted Y points for K-S test
Select this check box to use the plotted points (using median ranks or Kaplan-Meier) to calculate the difference between observed and model estimated probability for the Kolmogorov-Smirnov test, which is used by the Goodness of Fit Results window and the Distribution Wizard. If not selected, the non-parametric (empirical) values of the probability of failure are used as the observed value for the Kolmogorov-Smirnov test.
Use the Precision for Convergence Criteria slider to set the tolerance that will be used as the convergence limit (degree of accuracy) for mathematical iterations (e.g., a value of 4 indicates 1E-4 tolerance).
© 1992-2015. ReliaSoft Corporation. ALL RIGHTS RESERVED.