I try to estimate a dynamic factor model by ML estimation. The model has only one dynamic factor (inf_world, world inflation), and the data are quarterly inflation data for US, UK, JP, and DE. The problem is that the constant terms in measurement equations are not estimated properly but always estimated at the initial values. How can I fix it? Thanks!
dfm_inflation.mod (1.3 KB)
inf_g4.m (8.2 KB)
I still do not completely understand what is going on, but please try not using 0 as an initial value for the constants. There seems to be an identification issue. Using even 0.01 seems to result in a proper estimation.
Thank you for the reply, professor. I tried 0.01 rather than 0 as initial values for constant terms in the ML estimation, but then the estimation results for the constant terms in “oo_.mle_mode.parameters” just become 0.01.
I suspect there is an identification issue too, but the constant terms just capture the mean level of observation variables. So, I do not think there is an identification issue there.
I used that starting value together with mode_compute=5.