I think it depends on your real requirements (the aim behind your question) how expensive your task may be. This means what do you really want to know / to count - NULL's against what? The reason for it is that each field in Qlik is unique regardless how many times it was loaded respectively in how many tables it occur.
If it should really against each table you would need to apply such logic after each load or within the UI you would need a certain identifier-field for each table on which you could relate within your count-expression.
If I had such a task - of detecting and identifying issues with the data-quality - I wouldn't start with it in the script else in the UI by creating a pivot with $Table and $Field as dimensions and fieldvaluecount($Field) and sum($Rows) as expressions. The difference of both might then be your NULL's - whereby this is usually only part of checking data and often not the most interesting one because you might also have a lot of wrong and/or unexpected data.
To see the possible content of a field you could use two listbox - one with $Field and one with the expression: =$(='['&GetFieldSelections($Field)&']') and for any relations between your fields you could use a few tableboxes (if no unique id's are available for this kind of view you may apply some rowno() and/or recno() within your loads).
- Marcus
