Statistics for Exploring attrition bias: the case of the Khayelitsha panel study (2000-2004)