How to tell from SAS frequency distribution that something is off about my data? For example, I have 3 questions asking about the quality of a restaurant's services. Each question asks diners to rate from 1 to 5, where 1 indicates the worst quality and 5 indicates the best quality. If 100 people responded to the questionnaire and I read the data into SAS and make a frequency distribution of people's answers, how do I tell if something is wrong with the data just from the frequency distribution?
 A: There's two sorts of things to do here, depending on what is meant by "wrong".


*

*If there are values other than 1, 2, 3, 4 and 5 (eg "6", "0", or "fantastic") you know something is wrong.  This should be very obvious from the SAS frequency distribution, so obvious it might go without saying, but should be the first thing you check.  Similarly, the counts should add up to 100.  And so on.  These are the most basic things to check with what can go "wrong" eg coding or format errors.

*In the more general sense of going "wrong", assuming you basically have a 100 numbers from 1 to 5, there is no way you can tell just from the frequency distribution unless you have some kind of expectation.  Here the problem is that there are infinite things that can go "wrong", ranging from a small number of people getting the scale backwards (almost impossible to detect), through to the original data got scrambled and replaced with random numbers (could be detected if the random distribution of numbers is sufficiently different from a reasonable expectation).
For example, maybe you've done similar ratings with many restaurants before, and you find that most ratings are 3,4 or 5 and only a small number of people rate them as 1.  If for this restaurant you have 80 ratings of "1" it's surprising, but it doesn't necessarily mean the data are wrong.  It could be a really bad restaurant.  
There are ways of quantifying how surprising this current set of 100 numbers is compared to your expectation (based on previous ratings of this restaurant or other restaurants or just your vague prior expectations), but no statistical way of telling you whether the data are wrong or reality is just surprising.  In this case, statistics can tell you there's something to look into, but not what the cause is.
