Choose Language Hide Translation Bar
Highlighted
AGM
AGM
Level I

How do I generate actual vs predicted plots using the parametric survival platform?

Hi all,

 

I have been using JMP to analyse survival data (that contains right censored data), fatigue endurance data to be exact.

 

This data has been generated using a DOE matrix, and is investigating five independent variables.

 

What I would like to do is generate actual vs predicted plots with confidence intervals for the model that JMP comes up with from the parameteric survival platform.

 

I am new to the survival platform and dealing with data that contains censored data. We used to be able to use the fit model platform by transforming and this would give us everything we needed. However, I can't see how to do this using the parametric survival platform.

 

Another way that I could do it is if the parametric survival platform to save equations for the predicted values, 5 % CI and 95 % CI. Which I could then put into the graph builder to create a nice graph to put in a presentation.

 

Any help would be great.

 

Thanks

1 REPLY 1
Highlighted

Re: How do I generate actual vs predicted plots using the parametric survival platform?

If I understand your question correctly, you have basically provided the answer...

 

The fit parametric survival platform allows saving the quantile formula (from the menu under the red triangle). If you chose 0.5 as parameter you get an estimate after how many cycles 50% of your popluation have failed. I.e. a new column is added to the table having in every row the predicted 50% failure life given the conditions in that row.

You can then use the graph builder to plot the test data vs. the fitted failure 0.5 failure quantile which is basically an observed vs. predicted plot (see example). The dashed line is again the 0.5 percent quantile. By adding additional lines for e.g. the 0.05 and 0.95 quantiles you can add a confidence interval.

 

 

 

Graph Builder.jpg

Article Labels

    There are no labels assigned to this post.