Good afternoon!
I have a code in Python using Pykalman (https://pykalman.github.io/) where I create and I do not specify any parameters for the Kalman filter, just the number of dimensions of my observations. Initial values are initiated automatically (for example identity for the transition matrix) and then using EM algorithm, parameters of the Kalman filter are optimized. When it is done, we finally calculate the values predicted by our filter.
I struggle to do the same with R. I use the "dle" package (https://cran.r-project.org/web/packages/dlmodeler/dlmodeler.pdf). I then initiate my model and use the fit (with y a matrix with 84 columns and 2 lines):
mod=dlmodeler.build(a0 = c(1.142857143,-0.142857143), P0 = diag(2), P0inf = diag(2), Tt = diag(2), Rt = diag(2), Qt = diag(2), Zt = diag(2), Ht = diag(2), name='test')
dlmodeler.fit(y, model=mod, method='MLE')
Then I get this error:
Error in dlmodeler.build.function(model) : multivariate case is not yet implemented yet
Do you have any idea with this does not work and what should I modify in my methodology ?
Thanks.
