Question on mixed data sampling (MIDAS) regression

tigerbill · March 20th, 2012, 2:56 pm

Typical mixed data sampling (MIDAS) regression means the left and right side of regression can be at different frequency. I am wondering if there is econometric skill for the problem where even among the right side of regression, the variables can be at different frequency, for instance:Y = alpha + beta * X + epsilon,X is a vector of series, some in daily, some in weekly.Thanks a lot for your help in advance.

yuryr · March 21st, 2012, 8:22 am

gaussian processes is the answer (the trick is to use time difference as distance measure in cov function and only time)

tigerbill · March 21st, 2012, 9:40 am

thanks, yuryr, any reference?

yuryr · March 22nd, 2012, 12:48 pm

you'll be surprised: http://www.gaussianprocess.org/