When the responding sample doesn’t match the population in question (for example, if the population is equal parts Blue: 33.3%, Red: 33.3%, and Yellow: 33.3%, but the responding sample has more Blue than that — Blue: 45%, Red: 11.7%, Yellow: 33.3%), in order to make projections and assumptions about the larger population based on these responses, we want to make sure that we’re not under-representing the Reds or over-representing the Blues, so we “weight” the responses to match the “weight” of the population. In this case, each Blue response will count as 0.74 responses, Red responses will count as 2.84 responses, and Yellow responses will simply count as 1 response, resulting in a balance that matches the population: Blue: 45%*0.74=33.3%, Red: 11.7%*2.84=33.2%, and Yellow: 33.3%