Line 51:
Line 51:
What to do with the data of products that is clearly wrong? Maybe we should be more precise in how we handle and interpret data errors. I seem to distinguish two type of errors:
What to do with the data of products that is clearly wrong? Maybe we should be more precise in how we handle and interpret data errors. I seem to distinguish two type of errors:
−
* formal errors - these are based on rules we have established (Energy not > 3700 kJ, etc)'
+
* formal errors - these are based on rules we have established (Energy not > 3700 kJ, etc);
−
* interpreted errors - these are due to our understanding of the formal errors and are therefore more precise: kJ/kcal inverted, polyols missing, etc.
+
* interpreted errors - these are due to our understanding of the formal errors and are therefore more precise: kJ/kcal inverted, polyols missing, etc;
−
* outlier errors - data points that outside a formal envelope of a (cleaned) category.
+
* outlier errors - data points that outside a formal nutritional envelope of a (cleaned) category of products. We can subdivide these into values that are impossible (very large outliers) and strange (can not be correct);
+
+
Next question is what to do? Do we intervene, either by correcting or deleting?
== Graphing ==
== Graphing ==