When crop statistics are published by the source organizations, most will have some data points that are missing or seem like they may be erroneous. FEWS NET has standard practices to address these cases.
Missing data
Sources often document how they handle the missing data. Regardless of how a source organization documents missing data, FEWS NET represents it as follows:
-
0 (zed/zero): Used for the estimate of area planted, yield, and area harvested in cases where the source indicates no planting and/or production, no matter how the source indicates that. In many such cases, the source may only indicate this condition with a blank, which we normally treat as no area or production. We never use a “-” symbol to represent zed/zero, although some spreadsheets will use this symbol to display a value of zed/zero visually.
-
NC: Used when the source crop data is incomplete but it is clear that there should have been a value recorded. This is most often the case, for example, if the source provides a production estimate, but, for whatever the reason, no data point is given for area planted, area harvested, and/or yield. Many source orgs only provide one of the two area estimates (area planted and area harvested). In these cases, we insert an NC in the missing value’s place.
-
NA: Used if we are relatively certain that there was planting or production, but for whatever reason no estimate was provided by the source. This is sometimes the case when a source confirms it will not publish an estimate because there are too few producers of a crop, and to publish the estimate would violate the confidentiality expected by an individual farmer/producer.
Erroneous data
Clearly erroneous estimates are often found in source org reporting. In general, FEWS NET prefers to be conservative in assuming it knows when this has happened and how it should fix the error. When the cause of a clearly suspicious and likely erroneous estimate is not easily discerned, no change should be made and the source estimate should remain as it was reported.
When the likely cause of the suspicious estimate is easier to interpret, FEWS NET may consider “correcting” the estimate/s. As always, if possible, note should be made of corrections being applied to the source data. Cases where corrections may be made include:
-
A misplaced decimal
-
A mislabeled table
-
A weight or area measure being incorrectly, or not at all, converted