[Gretl-users] Confidence bounds on IRF

I am using the latest snapshot 64 bit version on Win8 64bit. I am having an issue the irf function. I get the err Matrix is not positive definite if I ask for bootstrapped confidence bounds, but there is no problem estimating the impulse response functions itself. This proble has only occurred with one of my sub-samples, so it might be hard to replicate. Any thought on what might be going wrong? Thanks, Logan