I want to cross validate various models to understand which one performs best. I have been referring to Rob J Hyndman's post here


But I am a little lost as to how to use it on daily data.

my daily time series data starts at 2015-01-01 and ends at 2017-03-31.

Specific questions include:

  1. What should k be in my case?

  2. In mae1 <- mae2 <- mae3 <- matrix(NA,n-k-12,12) instead of 12 will I use 365?

  3. In for(i in 1:(n-k-12)) will I use 365?

I have a very high level understanding of the code and I believe that is another reason why I cannot translate this to a daily usecase. Any help in clarifying the problem would be greatly appreciated!



Browse other questions tagged or ask your own question.