I am still learning jmp... Does the response have to fit a normal distribution or can it be log normal? The data is interval censored if that help or hinders answers. Any help would be appreciated.
A response's distribution is what it is...not much you can do about that. What you can control is the modeling procedures and techniques you use to analyze your empirical results. An assumption that is kind of crucial for the magic of ordinary least squares (we call it Standard Least Squares in JMP) regression to work properly is a normal distribution of errors for the response. Often times this assumption isn't one that matches a response set. Sounds like yours is an example of that case. But all is not lost...when the normality of errors assumption isn't met...usually the path is either:
1. Transform the response or;
2. Use an alternative modeling procedure or;
3. A mix of both 1. and 2. above.
From what little you've shared about your practical problem above...it sounds like you might be appropriate to fit some kind of degradation model? If that's the case you might want to check out JMP's Degradation and/or Destructive Degradation platforms. These platforms provide some easy means to transform a response, fit some classic or commonly encountered degradation models, censoring, including JMP's hallmark prediction profilers too.