How to specify priors for a 2 factor ALTI've searched JMP help and other locations, but I haven't been able to find an answer to this question. While I'm reasonably certain that I have the priors for Temp and Voltage correct in the pic below, I'm not certain how to fill in the intercept and the 1/β (scale/σ) parameters. Would I just add the two intercepts (β0) and average the 1/βs (σ)?
This is for a new customer who would like to do a 2 factor ALT. Any help would be greatly appreciated. I've attached the files I'm using so far.
Re: How to specify priors for a 2 factor ALTI recommend wait a little bit until JMP 18 gets released.
The existing ALT platform is deprecated.
In JMP 18, we recommend the new Constant Stress ALT Design platform, developed by @calking .
Re: How to specify priors for a 2 factor ALTHey @Jed_Campbell!
As @peng_liu pointed out, a better ALT platform is coming in JMP 18. The current ALT platform has some In the meantime, let's figure out how to answer your current question.
I don't believe adding the intercepts together would be a good idea. A better approach may be to try and determine what the levels of the other factors might have been in each experiment. The most reasonable assumption is that they are set to normal operating levels. For example, with the voltage, you could probably assume a reasonable operating temperature setting. Similar for voltage in the temperature experiment. If there is information about what these values actually might have been, then obviously that would take precedence. Doing this will allow you to then generate estimates of each slope and intercept that accounts for each factor being present.
For the sake of illustration, let's suppose the voltage is 8 in the temperature experiment and temperature is 20 C for the voltage experiment. I can then create a combined experiment table with voltage and temperature and then use the Fit Parametric Survival platform to estimate the parameters. For each of the factors, you can right click and choose the appropriate transformation, then move the transformed input to the Location Effects list. Once you run the report, you'll then have parameter estimates that you can use for priors as well as a covariance matrix you can use if you like.
I've attached a data table that shows what this looks like. If you have any questions, let me know!
Re: How to specify priors for a 2 factor ALTThanks @peng_liu and @calking! I'll be sure to mention that JMP 18 will have better ways for doing this. I also really appreciate the combined table approach, as I think it solves today's problem very well. I should be able to work the rest of this out in time to show the customer. If not, it's great having this resource.
