Closed mkao006 closed 10 years ago
Maybe an analytical data set should be built for the imputation. The following set of indicator may be used to impute yield.
Climate variables: (1) Temperature (2) Precipitation (3) Soil quality (4) Natural Disasters (5) Carbon dioxide concentration (6) Number of sunny days (7) Land
Management variables: (1) Irrigation (2) Fertilizer (3) Pesticides (4) Machinery/industrialization? (5) Genetics? Seeds
external data are no longer considered, see explanation in the methodology paper
External data such as temperature and precipitation can be easily downloaded from the world bank using the rWBclimate API.
Nevertheless, there are harmonization problem. For example, data for former and disputed country like USSR and Serbia and Taiwan does not exist. Furthermore, there may be missing values in these external data which we may need to impute before imputing the production domain.
One solution is to actually use multivariate imputation methods which will impute the production domain and the climate data at the same time. However, this will require further research and investigation.