I made a prototype of a weighted coarsen function for out workflow and added some tests.
For our purposes I had to restrict this to dataset where all variables have missing values in the exact same position, and where the weights (area usually) has zeros in those same locations, otherwise errors are raised.
I discovered some weird issue that I traced down to something to do with the precision of the weights, but could not reproduce with a simple example. I kept it in notebooks/jbusecke/issue_weighted_mean_precision.ipynb.
Some initial can be found in notebooks/jbusecke/dev_weighted_coarsen.ipynb.
I made a prototype of a weighted coarsen function for out workflow and added some tests.
For our purposes I had to restrict this to dataset where all variables have missing values in the exact same position, and where the weights (area usually) has zeros in those same locations, otherwise errors are raised.
I discovered some weird issue that I traced down to something to do with the precision of the weights, but could not reproduce with a simple example. I kept it in
notebooks/jbusecke/issue_weighted_mean_precision.ipynb
.Some initial can be found in
notebooks/jbusecke/dev_weighted_coarsen.ipynb
.