ABSTRACT
The idea behind weighting is simple: to make the respondents as similar as
possible to the original sample in terms of the distribution of some variables.
Suppose that in a sample of size 100, there were 60 women and 40 men.
Among the 60 women, suppose 40 provided information on income, while
only 10 among the 40 men provided income. Clearly, the respondents provid-
ing income have lost their original representation on gender. If the analysis
is restricted to respondents, then bias may be introduced, especially if the
income is correlated with gender. To compensate or correct for the loss of
representation, a weight of 60/40 = 3/2 = 1.5 is attached to each responding
woman and a weight of 40/10 = 4 is attached to each responding man. With
these weights, the respondents are now weighted back to the original sample
representation of 60 women and 40 men. The numbers 1.5 and 4 are called
nonresponse adjustment weights and the two cells formed based on gender are
the adjustment cells.