How to back-transform Mean Confidence Limits to the original response form

[#1]

Hello!

I built a multilinear regression model using natural logarithm transformation of my original response.

How can I back-transform Mean Confidence Limits JMP calculates for each particular predicted value to have it in terms of my original response form?  

Can I use next known relationship SD [ln(y)] ≅ RSD (y) to figure out PredSE confidence interval for my original response?

Thanks!

[#2]

Hi,

In Fit Model you can select your (original) response and then the red triangle menu next to Transform and choose log transformation. This will model the transformed response against the input variables and give predicted values and confidence intervals on the original scale.

Does that help?

Regards,

Phil

[#3]

Phil,

This is very useful indeed, but I wonder if JMP Mean Confidence Limits in log response space is still good for prediction of Mean Confidence Limits in the original response space? The mean in log space is not the same as in original space, although medians are the same, and It looks like JMP just does exp(Lower 95% Mean Limit) and  exp(Upper 95% Mean Limit), is this correct way to go? 

Thanks,

Konstantin