0
$\begingroup$

Just a general question that I couldn't find too much on.

What would be some good approaches to one step ahead forecasting of financial time series with mixed frequencies?

Often a lot of the available data influencing the price of say a stock or commodity is published at different frequencies, some daily, some weekly, some monthly etc. which in my head makes it tricky to use normal models for anything but the longest time interval.

Edit: found this paper which goes some way to cover the topic

$\endgroup$
  • 2
    $\begingroup$ state space models can deal with this. you go with the highest frequency, and the algorithm deals with missing data elements naturally $\endgroup$ – Aksakal Jun 10 '16 at 15:01
  • $\begingroup$ Interesting! got any good resources? @Aksakal $\endgroup$ – youjustreadthis Jun 10 '16 at 15:02
  • 1
    $\begingroup$ take a look at MATLAB ssm function help's Algorithms section. The first bullet's about missing observations. So, if you sample at monthly, and some data is quarterly, then it should be able to proceed assuming that not provided monthly data is "missing". $\endgroup$ – Aksakal Jun 10 '16 at 15:26
0
$\begingroup$

You can use MIDAS regressions to predict low-frequency target variable. See MATLAB toolbox for MIDAS regressions in many different contexts and actual data examples. https://nl.mathworks.com/matlabcentral/fileexchange/45150-midas-matlab-toolbox

$\endgroup$

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.