I'm trying to build a mixed frequency state space model containing high frequency dependent variables (weekly) and low frequency (monthly) independent variables. However, the Kalman filter does not seem to work when I want to forecast futrue weekly values as it cannot handle 'missing values' in the monthly regressors compared to the weekly dependent variables.
So for example my state space model look now like this:
@SIGNAL LOG(Y_weekly) = SV1+ SV2*LOG(X1_monthly)+SV3*LOG( X2_monthly)+SV5*LOG(X3_monthly)
@STATE SV1 =SV1(-1)+E1
@STATE SV2 = SV2(-1)
@STATE SV3 = SV3(-1)
@STATE SV5 = SV5(-1)
@EVAR VAR(E1) = EXP(C(1))
I know its possible interpolate the x_monthly regressors via the Kalman filter and then incorpate them in the weekly signal equation, however that is not the method I prefer to use.
I tried to read the article of Foroni, Guèrin and Marcellino (2015) - "Using Low Frequency Information for Predicting High Frequency Variables" which states that it should be possible with RU-MIDAS but to my knowlegde it is not an option in Eviews 10? Or with MF-VAR? However the article is a bit to advanced for me so I dont know completely understand it.
Do you have any advise how to approach this problem?
#I'm using eviews 10.
For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.
1 post • Page 1 of 1
Who is online
Users browsing this forum: Baidu [Spider] and 4 guests