Describe the bug
The default behaviour if no start points are given is to sample from a 0 centred normal with sd = 1 for each parameter. This will cause issues if the expected value for a parameter is far from this .
Expected behavior
Prior should be taken into consideration when sampling random effect starting points. This should include the prior$theta_mu_mean as the mean values and the square of the diagonal from prior$theta_mu_var as the standard deviations.
Describe the bug The default behaviour if no start points are given is to sample from a 0 centred normal with
sd = 1
for each parameter. This will cause issues if the expected value for a parameter is far from this .Expected behavior Prior should be taken into consideration when sampling random effect starting points. This should include the
prior$theta_mu_mean
as the mean values and the square of the diagonal fromprior$theta_mu_var
as the standard deviations.