Block candidate model. It produces an MLE that is non-unique, i.e.,
if estimating exponential series system, the MLE
for λ₁, λ₂, and λ₃ satisfies
λ̂₁ + λ̂₂ = λ₁ + λ₂
and
λ̂₃ = λ₃.
Usage
md_block_candidate_m3(md, qvar = "q", compvar = "t")
Arguments
- md
masked data
- qvar
column prefix for component probabilities, defaults to q,
e.g., q1, q2, ..., qm.
- compvar
Prefix-code for the component lifetimes, defaults to t,
e.g., t1, t2, ..., tm.
Details
This illustrates one of the conditions for the MLE to be unique:
(1) multiple components must not always show together. if they
do, then the best we can learn is that they satisfy some hyper-plane
constraint.