conditional mean and variance-co-variance matrix

I have 9 time series R, which is N*9 matrix. i would like to get conditional mean and variance for these time series. how can i write the code in matlab Under two different models? one is AR and GARCH, and another one is EWMA. i also need conditional variance-Covariance matrix, how to write the code under both of models.

thanks

Answer 1

[Adam Hug]

For the AR model, you can call the "estimate" function to generate an ARIMA/GARCH model object as well as the conditional variance-covariance matrix:

http://www.mathworks.com/help/econ/arima.estimate.html http://www.mathworks.com/help/econ/cvm.estimate.html

For an EWMA model, MATLAB has no builtin functionality that supports this. However, there are a few file exchange submissions that may help:

http://www.mathworks.com/matlabcentral/fileexchange/6503-covariance-tools-1-0a http://www.mathworks.com/matlabcentral/fileexchange/38302-kalman-filter-package

Both of these submissions contain tools for deriving the covariance matrix of an EWMA model.

[FEI]

Thank you for your suggestion, but i still confused. for example, i have data which contain weekly stock returns of 9 companies. Now, i want to construct a portfolio weights using mean-variance approach. in transitional method, the sample mean and variance-covariance is used. but, in my method, i would like to use conditional mean and variance-covariance matrix. how to estimate them in Matlab, if i use AR(1) and GARCH (1,1)? i know, the 'estimate' function can give me the estimated parameters of AR and Garch model. but it does not give me final conditional mean and variance matrix. such as use AR(1) and GARCH (1,1), model form are

the 'estimate' function just can give me parameters of model, but i need r and sigma. sure, if we estimated parameters, we can calculate the final r and sigma by hand. but, i perform a rolling-sample analysis, which means that i have to calculate them by Matlab code.

so, can you tell how to write the code?

Thank you.