Inverse Prediction

Use the Inverse Prediction tab to predict the time when the Y variable will reach a specified value. These times are sometime called pseudo failure times. Figure 6.14 shows the Inverse Prediction tab.

Figure 6.14 Inverse Prediction Tab

Enter either the Lower or Upper Spec Limit. Generally, if your Y variable decreases over time, then enter a Lower Spec Limit. If the Y variable increases over time, then enter an Upper Spec Limit.

For the GaAs Laser example, enter 10 for the Upper Spec Limit and click Go. A plot is produced showing the estimated times until the units reach a 10% increase in operating current. See Figure 6.15.

Figure 6.15 Inverse Prediction Plot

The Inverse Prediction red triangle menu has the following options:

Save Crossing Time

saves the pseudo failure times to a new data table. The table contains a Life Distribution or Fit Life by X script that can be used to fit a distribution to the pseudo failure times. When one of the Inverse Prediction Interval options is enabled, the table also includes the intervals.

Set Upper Spec Limit

is used to set the upper spec limit.

Set Lower Spec Limit

is used to set the lower spec limit.

Set Censoring Time

is used to set the censoring time. The plot updates to show the Censoring Time as a dotted vertical line. If Inverse Prediction Interval > No Interval is selected, observations that exceed the Censoring Time are displayed on horizontal lines starting at the Censoring Time. If Confidence Interval or Prediction Interval is selected, horizontal lines extend indefinitely to the right of observations whose upper limits exceed the Censoring Time. The Censoring Time is reflected in data tables constructed using Save Crossing Time and Generate Pseudo Failure Data.

Use Interpolation through Data

uses linear interpolation between points (instead of the fitted model) to predict when a unit crosses the spec limit. The behavior depends on whether a unit has observations exceeding the spec limit as follows:

–

If a unit has observations exceeding the spec limit, the inverse prediction is the linear interpolation between the observations that surround the spec limit.

–

If a unit does not have observations exceeding the spec limit, the inverse prediction is censored and has a value equal to the maximum observed time for that unit.

Inverse Prediction Interval

is used to show confidence or prediction intervals for the pseudo failure times on the Inverse Prediction plot. When intervals are enabled, the intervals are also included in the data table that is created when using the Save Crossing Time option.

Inverse Prediction Alpha

is used to specify the alpha level used for the intervals.

Inverse Prediction Side

is used to specify one or two sided intervals.