Idea: when data contains age patterns where implied sensitivity can be assumed smooth, it might be practical to only evaluate the decomp at selected points, back out the sensitivity, and do a spline interpolation to all age groups. But if there are multiple age patterns we need a way to split, like a factor or something (e.g. causes of death, or different state transitions). So maybe best use this trick if post-processed; that way one can be picky about the interpolation method? To do this internally would require info in the ... arg.
Idea: when data contains age patterns where implied sensitivity can be assumed smooth, it might be practical to only evaluate the decomp at selected points, back out the sensitivity, and do a spline interpolation to all age groups. But if there are multiple age patterns we need a way to split, like a factor or something (e.g. causes of death, or different state transitions). So maybe best use this trick if post-processed; that way one can be picky about the interpolation method? To do this internally would require info in the ... arg.