How to clean false values of a variable?

[PierreYvesILY]

Dear SAS experts,

I have a set of variables and among them name and surname: NACHNAME and VORNAME. These are character variables length $35.

Some of their values are not correct: for instance, you can find :

• cccc
• ccc
• ?0???
• !-§4$
• Test

Is there a method to clean them (delete the lines, where NACHNAME or VORNAME have this kind of irrelevant value)?

Thanx,

regards

PY

[PaigeMiller · Accepted solution]

What are the rules that determine if a value is not acceptable?

[PierreYvesILY]

We are in Germany, so (same rules NACHNAME and VORNAME):

- all names must be written with Latin letters, of any european language (not possible to select this I guess)

- in a name, you can have as many words as you need: 'Du Taxi du Pouet de la Valse Folle' is correct

- the sign - between 2 words is accepted, as well as a space, or both together (people also type errors), and also '

- accents on voyels are OK : ` ´ ^ ~ ö ä ü

- the _ will be tolerated - the following signs determine in all cases a false name: ? , = ; ( ) / & % $ § ! " * + # @ < > > | €

- such test cases names as : cccc, ccc, Test, test... will be eliminated (I noticed some)

My purpose here is to be helped in the method to achieve this goal, I would like to progress and understand how I can do.

[PaigeMiller]

false_name=findc(name,'?,=;()/&%$§!"*+#@<>>|€')>0;

As far as the other types of false names, you would need to create general rules that could be programmed, such as if the same letter appears 4 times consecutively, that is a false name. Naturally, there are many such general rules that would have to be defined and then programmed. 

[PierreYvesILY]

I implemented this solution with success.
Thank you.

[PGStats]

Adding a rule rejecting repeated characters:

data want;
set have;
invalid = 
	findc(name,'?,=;()/&%$§!"*+#@<>>|€') > 0
		or
	prxmatch("/(\S)\1{2,}/", name);
run;