How to solve the problem of low neff value

Now I have some problems about how to solve the problem of low neff value. When I do a simulation with a program, the estimate is reasonable. But when the program is used to estimate actual cases,the estimate is unreasonable,now what I should do?
And there is also a question, Four chains are used for mixing, three of which converge,but one had more deviation,then I should do for it ?

It indicates that the model might not be entirely good at estimating your parameters. This can be due to many things. Can you provide a minimum working example?

1 Like

Sounds like (i) your model converging or being identifiable is dependent on the data which may be bad in itself (ii) your synthetic data doesn’t look like the real data.

Generically you could try run your model for more iterations and use a higher delta adapt value but I’ve rarely seen that help in reality.