Propagating error of mean for use in ANOVA

[#1]

I have used an external program (MATLAB) to fit a logistic curve to 84 sets of y-response vs. time data (n=3 for 28 different factor combinations). My logistic model is a 4-parameter model (A, B, C, D) and I get an upper 95% confidence interval and a lower 95% confidence interval for each parameter (see column headers below to understand what data is outputted by MATLAB). I would like to use JMP to output the mean of each parameter for my 3 replications, but I should also 'propagate' the error of each parameter to get the final error of the mean, correct? I then plan to do an ANOVA to compare these parameters with respect to the factors I have varied. How can I ensure JMP takes account the error for each parameter of each data entry?

[#2]

Bootstrapping the parameter estimates would be one way to go.

I've used this approach too: Simulating Multi-Step Processes and the Propagation of Error to the Final Response 

[#3]

I agree... JMP needs a way to do this... So far, I don't see a way to do it... However, I have several ideas of how they can program the thing in. The easiest way would be, make a new "Error Column" with the same function to propagate, and  program a matching strdErr for each variable in the Function - It can just be side-by-side style... Then, the result of the function can be the propagated error... There would be two columns, one for the actual function value, and another for a stdErr propagated.

Maybe even easier would be to just make a Column Property that's called "stdErr" that has a property that matches with another column it's associated. Then, JMP could just Universally apply the propagated error, when the errors are associated with every member of the function.

In any case, this is pretty huge.. JMP also lacks the ability to easily perform ANOVA when you've got only Avg, StdErr, and N ... There's a way to do it, but it really should be easier.... All analysis is fitting models on data, pulling the parameters out (with their stdErr) and doing some statistical comparison... JMP, as far as I know, severely lacks the ability to propagate errors... 

It really seems simple to program, and base it directly on the function at hand, and apply symbolic calculus to the thing... Lots of software do that... Once you've got that... You're basically done... JMP should really add this in!

[#4]

Up in the blue bar at the top, click on the JMP Wish List Button, and let the developers see what you're looking for.