Extrapolation of yhat


New Member

I am using non-linear regression ('nl' function, with independent variable of age) to estimate coefficient with the least sum of the square value (fitted non-linear curve).

Later, I was using 'predict yhat' to predict the y-value for my samples with different age groups. However, the samples I had only range from age 10-20 years and I would like to extrapolate and estimate the yhat value for larger range, exp: 0 to 50 years, anyone knows how can I do this?

Thanks for your help.
Last edited:


Active Member
This is a bad bad idea from the point of view of statistical methodology. Any statistical model should be used for predictions only close to the range of the training data, unless the model puts extra rigidity in the tails to avoid explosive behavior (e.g. models based on extreme value theory, natural splines, etc).