do you believe in grounded way of doing things or are you too sure about your models?

Hi 

There is a nice chapter in Andy Filed book 'discovering statistics using SPSS' on the use of nonparametric tests (in case you are using other statistics software Andy Field has another book for R, 'discovering statistics using R' with deep detailed information). And yes, Mann-Whitney U test can be used to determine differences among two independent means. But there are some drawbacks to this type of analysis, you may wish to consult Andy's book suggested earlier. Best. 

The first task would be to specify and then quantify the salient or key linguistic features and patterns known to be produced by other speakers of the dialect or variety being studied, in this instance African American English. Once the key linguistic features and patterns are identified, a Dialect Density Measure (DDM) can be calculated.  For additional information on coding these features in African American English, see Van Hofwegen, Janneke and Walt Wolfram. 2010. Coming of age in African American English: A longitudinal study. Journal of Sociolinguistics 14: 427-455.

The literature mentions 3 ways of calculating DDM.  They are: 1) number of utterances with one or more pattern divided by the total number of utterances produced by the speaker; 2) number of patterns produced by the speaker divided by the total number of utterances produced by the speaker; 3) number of patterns produced by the speaker divided by the total number of word produced by the speaker.

In an article "Methods for characterizing participants' nonmainstream dialect use in child language research" that appears in the Journal of Speech Lang Hearing Research. 2002 Jun;45(3):505-18, and focuses on dialect use in child language research Oetting & McDonald demonstrate that the 3 approaches are highly correlated.