regression line in proc gplot

[mikepk]

Hi all,

I'm trying to plot and fit data to a power function (y=ax^b) using proc gplot. The interpol option in the symbol statement allows you to define a type of regression (eg: rl=linear, rq=quadratic etc) but I don't see the power function as being one of them.

Is there perhaps another way to do this? I can easily do this in Excel as the trendline option has the power option........but need to use SAS to do it.

Thanks!

Mike

[GraphGuy]

Do you have some sample data you could post up?

[PaigeMiller]

Here's how I normally do this.

1. Augment the raw data with new observations that contain x-values that cover the range of your data and are spaced closely together. For example, if your data range is 0 to 10, I would add values 0, 0.01, 0.02, etc. The Y values are missing for these new observations.
2. Perform the regression on this augmented data set and save the predicted values
3. Plot the original data, and overlay the predicted values for the augmented data. At this point, you can simply connect the predicted value points linearly, that will be close enough for graphical purposes.