Hi,
I will be on holiday next week and even the documentation and tests of the function are not finished ('Draft' label in the title), @cchou and @cdelgado can report here their comments on:
- is the cross-validation correctly applied?
- is the NA behavior working as expected?
- are the dimension name parameters correctly working?
I would say this merge request is not aimed at exploring the behavior of all options that qmap package allows to apply. For instance:
- if you get zeros in your result, read the doc of qmap because there is a parameter 'wet.day' that can be set as FALSE to avoid this.
- if the adjustment fails, check if you select an appropriate distribution/method for your data. To understand all this, you need to read qmap or go to the scientific publications cited in the documentation.
The new parameters are:
-
sdate_dim
(default 'sdate'): dimension in which cross-validation is applied -
memb_dim
(default 'member'): if this dimension is not in 'obs', it will be internally created with length 1 -
window_dim
(default NULL): this dimension can be created by the user beforehand. It will be useful in the case of daily values when a window of time steps around each forecast time wants to be used for the adjustment. It could be part of obs and exp, but not in 'exp_cor' (does this makes sense to you?) -
na.rm
(default FALSE): logical indicating if the NAs should be removed or not for calculations
I attach again a document with some tests that may be later transform into unit tests testQM.R
I have also created an example of how to create the window dimension, please, you can provide feedback on this too, it could be converted into a vignette also caseQM.R. Maybe @jramon wants to check it too.
I am not sure if I am missing something...
Have a nice week!
Núria