I am able to generate GAM models using the package mgcv in R. However, I need to write out the equation explicitly. How can I do that? The command 'coeff' gives me intercepts and coefficients, but I do not know what these coefficients really mean!
GAMs do not have an explicit form that can be nicely written down with algebra. Instead, they "smooth" over the data using kernels that determine the wiggliness the fitted model. I'm sure you could numerically probe the fitted model so that you can get the output for any input, but I don't think there is much value in attempting to write down the function.
You are right that they do not have an explicit form that can be nicely written. I need to use the model in a C# program where I can do new predictions within the code there. It is part of a mesh deformation software that the model will enter. So, I do need explicit equations that I can use. I can see that coef(out) gives me 9 coefficients for each predictor, but these coefficients should be part of the spline functions; and the spline functions are not laid out in any manual. That is a pity as it reduces the utility of these GAM models to within R or exporting R outputs as a data file for use elsewhere, which is cumbersome.
No explicit form as it follow to each point you have. You can represent you gam model using graphical approaches.
The use of natural spline will help in smoothing each edge of the curve in gam. If the smoothing parameter value is very big the gam line tend to be over smooth and you might get straight line. Model selection need to be carried out. I strongly belief for gam the graphical presentation give more information than trying to get the equation that can representing the gam line.
Ok, good. But, I do need explicit equations as they need to be inserted elsewhere.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R