How do I estimate a state-space model using ParamMap?

I created an implicit state-space model Md1 with a function ParamMap.

When I use the command EstMd1=estimate(Md1,Y,parms0) where Y is my observed data and params0 is a vector containing my initial values for the unknow state I get the error "Length of the state type vector must agree with the number of states."

Is the bolded command the correct syntax for estimating implicit state-space model?

(note that when I estimate the model by creating the state-space matrices explicitly in the main program it works and so I know that the number of states agrees with the length of the state vector)

function [A,B,C,D,Mean0,Cov0,StateType] = ParamMap(parms0)

var1=exp(parms0(2)); % Positive variance constraint

var2=exp(parms0(3));

A=[1 0; 0 parms0(1)];

C=[1 1];

B=[var1 0;0 var2];

D = [];

Mean0 = 0.5;

Cov0 = 10;

StateType = 0;

end

Answer 1

[Shantanu Dixit]

Hi Ed, the error you are encountering is due to the fact both 'Cov0' and 'StateType' are defined in 'ParamMap'. When implicitly creating a model by specifying 'ParamMap', both 'Cov0' and 'StateType' are empty vectors [ ] and are specified by 'estimate' function after estimation.

Refer to the below sample code

% dummy data
Y = randn(100, 1);
parms0 = [0.9, log(0.1), log(0.2)]; 


function [A, B, C,D, Mean0, Cov0, StateType] = ParamMap(parms0)
    var1 = exp(parms0(2)); 
    var2 = exp(parms0(3));
    A = [1 0; 0 parms0(1)];
    C = [1 1];
    B = [var1 0; 0 var2];
    D = [];
    Mean0 = [0.5; 0.5];
    Cov0 = []; 
    StateType = []; 
end

% Create the implicit state-space model
Md1 = ssm(@ParamMap);

% Estimate the model parameters
EstMd1 = estimate(Md1, Y, parms0);

Refer to the following MathWorks documentation for better understanding:

• estimate: www.mathworks.com/help/econ/ssm.estimate.html
• ssm: www.mathworks.com/help/econ/ssm.html