Are the data really in accord with (a) the assumed distribution and (b) the assumed model? Clues to (a) may be obtained from a plot of the ordered observations against some simple function of their ranks, preferably on probability paper appropriate for the distribution assumed.