Muriel Vasconcellos (asker) Nov 25, 2014:

@Charles & Donal I agree with your explanation. As I move forward, I'm encountering the same idea elsewhere in the document, but expressed a little more clearly.

Charles Davis Nov 25, 2014:

@Muriel I am a duffer at statistics, so my opinion counts for little here, but as I understand it, it doesn't mean "pick any one of the variables being used and calculate the highest percentage of outliers for that particular variable", but rather: "out of all the variables being used, find the one that has the highest percentage of outliers and calculate that percentage". In other words, "any", but not "any one".

Muriel Vasconcellos (asker) Nov 25, 2014:

See a later passage I have now encountered a passage where it clearly states that they are looking for the highest percentage of errors for each type of variable, which makes more sense. Do you think this applies to my question as well?

Además, es aconsejable **calcular el mayor porcentaje de registros sin código por tipo de variable** (si es que se realiza la codificación de alguna variable), el cual no debería sobrepasar el 20%. Esto corresponde a casos que no se pudieron codificar debido a que, por ejemplo, la descripción de la variable abierta es insuficiente o ambigua. Por eso es fundamental que el personal responsable de la captación de los datos del registro administrativo esté formado en el uso del clasificador correspondiente, de modo tal que pueda detectar y corregir estos casos inmediatamente.