search

cran / quarks

:exclamation: This is a read-only mirror of the CRAN R package repository. quarks — Simple Methods for Calculating and Backtesting Value at Risk and Expected Shortfall
1 stars 2 forks source link

Rolling forecasts and backtesting #1

Open MislavSag opened 3 years ago

MislavSag commented 3 years ago

Hi,

Thanks for the nice package.

There is an nwin argument to rollcast function. On first I thought this argument is used for rolling window, but when I check the code, it seems to me it is just used to subsample the whole sample. For example, if there are 1000 observations and I set nwin to 100, only 100 observations would be used and other discarded (if we ignore nout argument).

What I am interested in is rolling windows, that is backtesting. For example, I would like to use 100 tolling windows, than forecast for next time step (minute or hour) or multiple steps and check in the end how good are the forecasts.

I usually use runner package for this kind of tasks, but just want to be sure quarks package doesn't already have this kind oh feature.

gaborcsardi commented 3 years ago

Hi, this is a read only mirror of CRAN, please see the package authors and possibly URLs in the DESCRIPTION file. Thanks!