this refers this paper from NY fed on a regression based approach to TextI am trying to replicate the results in R. While this works very good for estimating the rates in the affine model (eq 25 to 27 in the paper), I am having trouble estimating the "risk neutral" rates (and hence the term premium). As the paper mentions (page 10, last paragraph), putting the price of risks as zero generates the risk neutral rates. However I am getting correct real world rate, but very off risk neutral rates by putting lambda0 and lambda1 to zero running the recursion in eq 25 to 27. Any pointer what I am getting wrong. Thanks vm.