Skip to content

Longitudinal logistic regression with continuous-time first-order autocorrelation structure

1 message · Ben Bolker

#
Don't use (time_hours|id) ... that will expand to a random effect with
a full, unstructured covariance matrix term.
If you have t distinct times measured, you'll end up with t*(t+1)/2
parameters to estimate.  Try (1|time_hours)
(and probably also include (1|id))

On Wed, Feb 28, 2018 at 8:44 PM, Dennis Ruenger
<dennis.ruenger at gmail.com> wrote: