|
How can I tell if a transformation is required for my response
in DOE++ and, if so, which one I should use?
The Residual Probability plot is a simple way to evaluate the
fit of the model in DOE++. If all points fall more or less on the line,
as shown next, the model is good.
If a noticeable pattern is
present in the plot, as shown next, it may be appropriate to
apply a transformation to the response.
The Box-Cox Transformation plot
can help determine what transformation, if any, should be
applied. The plot shows the sum of squares of the residuals
plotted against lambda. The value of lambda at the minimum point
of this curve is considered the "best value" of lambda, and
indicates the appropriate transformation, which is also noted by
default in the lower title of the plot.
[Click
to Enlarge]
To apply a transformation to
the response, click the Transformation icon in the Design
tab or Analysis tab Control Panel.
On the General page of the
Transformation window, select the response you want to work with
from the Response drop-down menu. (Note that when the
Transformation window is accessed from the Analysis tab Control
Panel, the choice you make will be applied only to the response
currently selected in the Control Panel. The Response
drop-down menu will be disabled.) You can then select a
transformation from the Transformation drop-down menu.
The equation for the transformation will be displayed at the
bottom of the window and text fields for any required inputs
will be available in the Parameters area, as shown next.
Note that the information
displayed in the Box-Cox Transformation plot is also available
on the Box-Cox Transformation page of the Transformation window.
Click the Calculate button in the Box-Cox Transformation
Results area. The Box-Cox Transformation Results area will then
display the best lambda value for the response, including the
two-sided confidence bounds at the level determined by the risk
level (alpha) you have specified, which helps in selecting an
appropriate transformation. The recommended transformation, if
any, will be displayed in the Recommended Transformation area,
along with any recommended parameters, as shown next.
You can then return to the
General page of the Transformation window to implement the
recommended transformation.
Why is "Pnz" sometimes displayed in the results area of the
Standard Folio Control Panel in Weibull++?
Pnz stands for "percent non-zero." When your data set contains
failures at time 0, this figure is automatically displayed in
the results provided by
Weibull++.
Percent non-zero is calculated
as follows:
Pnz = 1 -
(Number of failures at time 0) / (Number of all entries – Number
of suspensions at time 0)
The suspensions at time 0 are
ignored and do not count in the analysis.
Percent non-zero is also
displayed by default in the lower title of plots, along with the
calculated parameters. It is denoted by Z. |
