Negative Data Points when using MNRFIT function

I am trying to use MNRFIT to create a logistical regression for pseudo-zscores of neural signal data and the targets that they are reaching. Some of the pseudo-zscores I am using are negative values and MNRFIT shows this error because of it:

Error using linsolve Matrix must be positive definite.

Error in mnrfit (line 242)

            bcov = linsolve(hess,eye(size(hess)),struct('SYM',true,'POSDEF',true));

I was wondering if there is anything I can do to fix this problem. If there is any way to adjust data without affecting it. I m not sure if shifting all the data points by the minimum to the right would be a appropriate solution since it may affect the regression.