See dynare.org/news/release-of-dynare-4.2.5 for more details.
After updating from 4.2.4 to 4.2.5 I receive the error message below with a file that used to run fine under the old version. Does anyone have an idea how to fix this?
Error using inverse_gamma_specification (line 121)
inverse_gamma_specification:: Failed in solving for the hyperparameters!
Error in set_prior (line 224)
[bayestopt_.p6(k(i)),bayestopt_.p7(k(i))] = …
Error in dynare_estimation_init (line 124)
[xparam1,estim_params_,bayestopt_,lb,ub,M_] = set_prior(estim_params_,M_,options_);
Error in dynare_estimation_1 (line 37)
[data,rawdata,xparam1] = dynare_estimation_init(var_list_);
Error in dynare_estimation (line 62)
Error in ad4 (line 301)
Error in dynare (line 120)
This occurs when the variance of an inverse gamma prior is too small or too large in comparison with the mean. In order to find the natural parameters of the inverse gamma distribution from the mean and variance, Dynare solves a non-linear problem. In earlier version, we didn’t check the accuracy of the result and let pass inaccurate solutions
In version 4.2.5 the accuracy of this computation is checked and if the result is inaccurate, an error is sent.
The solution is to modify the ratio mean/variance in the prior. Remember that it is often better to calibrate a parameter than to use a very small variance for a prior. Note also, that using in infinite variance for the inverse gamma pdf rahter than a large variance makes the above computation easier.