search

helske / bssm

Bayesian Inference of State Space Models
40 stars 14 forks source link

Question about nlg_ar_exp #31

Closed shubingtang closed 2 years ago

shubingtang commented 2 years ago

I checked ‘cpp_example_model’ function and found that ‘nlg_ar_exp’ does not apply to my model, I wonder whether my model is supported by bssm package , my model is as follows:

Q{t}=exp(y{t})+exp(y{t-1})+exp(y{t-2}) y{t}=tau{t}+c{t} tau{t}=tau{t-1}+d{t-1}+eta{t} d{t}=d{t-1}+V{t} c{t}=a{1}c{t-1}+a{2}c{t-2}+epsilon{t}

Many thanks!

helske commented 2 years ago

I'm not sure I follow your equations, what is Q? If you can write your model in a way described in the bssm paper/vignette/docs (either in R for ssm_ulg etc or in C++ for ssm_nlg), then in principle the model is supported, but if I guess correctly your formulation in terms of states and observations, I think there will be issues with the identifiability.

shubingtang commented 2 years ago

Q is observations,I've always wondered if the model is set up correctly, thank you a lot.