Hi,
I’m using the package “DynamicFactor” from statsmodels.tsa. I’m having sort of problems when trying to fix a parameter. The code that I have done is the following:
model_restr = DynamicFactor(endog_m, k_factors=factores, factor_order=2,error_cov_type=‘diagonal’, enforce_stationarity= True )
with model_restr.fix_params({‘loading.f2.Servicios’: 0}):
res_restr = model_restr.fit(disp=False)
The issue is that, when I estimate the model without constraints via (I estimated the model twice):
model = DynamicFactor(endog_m, k_factors=factores, factor_order=2,error_cov_type=‘diagonal’, enforce_stationarity= True )
initial_res = model.fit(disp=False)
res = model.fit(initial_res.params, disp=False)
All the loading factors are significant for the model, except the one that I fixed (‘loading.f2.Servicios’). However, onceI fixed the parameter, all the loading factors become non significant for the model. It seems very strange for me, because I expected the contrary. Maybe I’m skipping something.
Thanks in advance!