10
$\begingroup$

In a previous question, I inquired about fitting distributions to some non-Gaussian empirical data.

It was suggested to me offline, that I might try the assumption that the data is Gaussian and fit a Kalman filter first. Then, depending on the errors, decide if it is worth developing something fancier. That makes sense.

So, with a nice set of time series data, I need to estimate several variable for a Kalman filter to run.

(Sure, there is probably an R package somewhere, but I want to actually learn how to do this myself.)

$\endgroup$
6
$\begingroup$

Max Welling has a nice tutorial that describes all of the Kalman Filtering and Smoothing equations as well as parameter estimation. This may be a good place to start.

$\endgroup$
1
$\begingroup$

The usual method is to use Maximum Likelihood Estimation. Basically, you need a Likelihood function and then run a standard optimizer (such as optim) to maximize your Likelihood.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.