The current delayed rejection in #136 is not complete because it assumes that the proposal distribution does not depend on the already rejected samples. The publication at https://arxiv.org/pdf/0904.2207.pdf gives a nice overview of how the acceptance probability for a second, third, ... sample has to be adjusted to take into account the delayed samples. We should fix this in our implementation.
Reactions are currently unavailable
The current delayed rejection in #136 is not complete because it assumes that the proposal distribution does not depend on the already rejected samples. The publication at https://arxiv.org/pdf/0904.2207.pdf gives a nice overview of how the acceptance probability for a second, third, ... sample has to be adjusted to take into account the delayed samples. We should fix this in our implementation.