[x] 2) clarify the logic of fixedpars option in Ardo's example (line 68). Why 8? How does it differ from constraint?
[x] 3) For the cascading estimation technique, in which the estimated transitions are used as starting values, how does one obtain the new Q matrix? Should we use qmatrix.msm(model) and then replace some elements by 0 to match the initial Q?
[x] 4) What series of model specification should be tested in search of the "correct" model to be used by all studies? See #27 for the latest (version 3) specification.
Some ideas
[x] should we use a common specification object? (e.g. model_specification, version 3)
[x] scripts for single model testing and examination. Might be of use to organize your workflow.
Discussion topics
E
(see https://github.com/IALSA/ialsa-2016-amsterdam/issues/29#issuecomment-244223270)fixedpars
option in Ardo's example (line 68). Why8
? How does it differ fromconstraint
?Q
matrix? Should we useqmatrix.msm(model)
and then replace some elements by0
to match the initialQ
?Some ideas
model_specification
, version 3)