Question 200:I've calculated the confidence intervals about a regression line. My data is log transformed and has high R-square value for what is called "power" equation in Excel. My CI's are narrow at the top of the graph and very wide near the bottom. Can I use the "standard" method used to calculate CI about a linear regression curve for the log transformed data and apply it to the power equation? My CI calculation used is
CI = t*SYX*SQRT(1/n+(X-XAVG)^2/SS
When I have a CI envelope that is narrow at top and very wide at the bottom of the regression line what does that mean?
Interestingly, when the R-square value is low the CI plots about the regression line make a much more "palatable" envelope. This may be because both the curves generated from linear and power equations have slopes and similar r-squares.