Closed gowerc closed 1 year ago
Recommendations from @danielinteractive were to check the following:
Phi with true value of 0.1
for 800 / 1600 / 4000 patients respectively:
variable mean median sd mad q5 q95 rhat
link_lm_phi 0.132 0.133 3.49e-2 3.46e-2 0.0756 0.188 1.00
link_lm_phi 0.0881 0.0886 2.48e-2 2.46e-2 0.0458 0.127 1.01
link_lm_phi 0.117 0.117 1.57e-2 1.49e-2 0.0898 0.142 1.00
Seems to be working exactly as expected now will close
Currently it looks like the link parameter to the random slope model isn't working. As in its credible intervals are massive even with reasonably large amounts of data. This may just be the nature of the model but it may also indicate a bug in the code. Needs to be investigated and resolved.
example code: