In Weibull++, how can I use the Specify Points tool
to mitigate the effect of outliers in my data set?
One way of treating outliers in Weibull++ is to use the Specify Points window. You can mark outliers in
the data sheet by applying a subset ID. Then, you can use the Specify Points tool
to exclude those
data points from the calculation of the regression (the median ranks calculation will still include
all data points). This will mitigate their influence on the analysis.
NOTE: The Specify Points tool will only work if you choose to use rank regression (least squares) for your analysis.
First, enter your data in a standard folio and then identify each data point with a subset ID, making sure
that the data points you wish to mark as outliers all have the same subset ID. If the data sheet is
configured to use Maximum Likelihood Estimation (MLE) by default, toggle the selection to
Rank Regression on X (RRX), or select the option
from the Analysis tab.
Next, choose Data
> Specify Points to open the specify points window. The tool displays a list of all the subset IDs
that have been defined in the data sheet and a list of those that have already been selected. To select one
or more IDs, you can a) click and drag the ID from one column to the other, b) doubleclick the ID, or c) use
the Select All Available or Clear All Selected buttons. Select the ID(s) that you want to include in the
calculation of the regression. The ID(s) that are not selected (i.e. the outliers) will be excluded.
When you close the window and return to the standard folio, you will see the Points Specified indicator
in the control panel and the parameters will automatically be recalculated based on the subset ID(s) that
were selected.
The following plot illustrates the effect of excluding the outliers from the calculation of the
regression. The black line (Folio2) shows the analysis that treats all data points the same way. The
blue line (Folio1) shows the analysis using the Specify Points tool to mitigate the influence of the
outlier data points. As you can see, the analysis that excludes the outliers from the calculation of
the regression provides a much better fit for the data points.
In Lambda Predict can I include parts that are not in my selected
reliability prediction standard?
Yes. In Lambda Predict 3, you can include information about parts that are not included in a
selected reliability prediction standard. You do this by adding an External Component to your prediction.
Choose System Heirarchy > Add Same Level Item (or Add Next Level Item)
> Component/Block. From the Select a Component window, select
an External component. You can then set
the properties of the component, such as the name, failure rate, quantity and adjustment factor.
For MILHDBK217F systems you also define the number of pins.
External components are available for all reliability prediction standards. The General properties
group includes the same properties in all standards. The remaining properties groups vary slightly
depending on the standard.
Note: For Telcordia SR332 Issue 2 systems, external components always display a dash in the Failure
Rate Upper Bound(t=inf) column in the System Hierarchy panel. This is because while the bound calculation
takes into account the base failure rate of the external component, it assumes that the external
componentâ€™s failure rate has no associated variability.
