Info

此问题已关闭。 请重新打开它进行编辑或回答。

Error while fitting a time varying state space model

1 次查看（过去 30 天）
tirupam goel
tirupam goel 2015-7-17
关闭： MATLAB Answer Bot 2021-8-20
I am trying to fit a time varying state space model. Essentially, I have mixed frequency data, which entails to a time varying observation matrix. I think I have coded up correctly (looking up http://www.mathworks.com/help/econ/estimate-a-time-varying-state-space-model.html). However, I am getting the following error:
Length of the observation vector in period 2 must agree with the number of rows in matrix C.
I am attaching the code and data if someone would like to have a look at that.
% Function that develops the model
function [A,B,C,D,Mean0,Cov0,StateType] = ParamFun_monthly_small(p,T)
% A
A1 = zeros(8,8);
A1(1,1) = p(1); A1(1,2) = p(2);
A1(2,1) = 1;
A1(3,3) = 1; A1(3,4) = 1;
A1(4,4) = 1;
A1(5,5) = 1; A1(5,6) = 1;
A1(6,6) = 1;
A1(7,1) = p(3); A1(7,2) = p(4);
A1(8,8) = 1;
%%B (use exponential of variance parameters to ensure non negativity)
B1 = zeros(8,8);
B1(1,1) = exp(p(5));
B1(4,4) = exp(p(6));
B1(6,6) = exp(p(7));
B1(7,7) = exp(p(8));
B1(8,8) = exp(p(9));
%%C
C1 = zeros(3,8);
C2 = zeros(2,8);
C1 = {[1 0 1 0 0 0 0 0;
p(10) p(11) 0 0 1 0 0 0;
0 0 0 0 0 0 1 1]};
C2 = {[p(10) p(11) 0 0 1 0 0 0;
0 0 0 0 0 0 1 1]};
C_augment = [C1; C2; C2];
%%D
D1 = zeros(3,3);
D2 = zeros(2,3);
D_augment = [D1; D2; D2];
Mean0 = [];
Cov0 = [];
StateType = [0 0 2 2 2 2 0 2];
A = A1;
B = B1;
C = repmat(C_augment,T/3,1);
D = 0; %repmat(D_augment,T/3,1);
end
% Main code that fits the model
data = xlsread('U:\SGN_Tirupam_Current\local-natural-rate\new\Data\Interwrk\monthly_data_model.xlsx');
% year quarter gdp cpi ea jc u eu ue un ne nu eep ea_1995q1 ea_1998q1 ea_2000q1 jc_1997q3
%Order of variable for the model = y eu ue un ne nu eep u
names = {'gdp' 'u' 'cpi'};
obs = [data(:,3) data(:,7) data(:,4) ];
T = length(obs);
obs = obs(3:T-1,:); %so that first row has all observations
Mdl = ssm(@(params) ParamFun_monthly_small(params,T));
load phi.mat phi omega beta;
% remove coeffs of pi,u,ea,jc
% phi([2,3,10,11],:) = [];
p0 = zeros(11,1);
p0(1) = phi(1,1);
p0(2) = phi(1,2);
p0(3) = beta(1);
p0(4) = beta(2);
p0(5) = log(0.01);
p0(6) = log(0.01);
p0(7) = log(0.01);
p0(8) = log(0.01);
p0(9) = log(0.01);
p0(10) = omega(1);
p0(11) = omega(2);
%no measurement error as of now - we can introduce it in u OR ea OR both
%depending on how much can be identified
tic
options = optimoptions('fminunc','MaxFunEval',2000);
[m,estParams,~,logL,Output] = ...
estimate(Mdl,obs,p0,'Display','full','Options',options);
toc
state = smooth(m,obs);
end

此问题已关闭。

回答（0 个）

此问题已关闭。

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by