1. Weight matters
This paper deals with weighting, its function in statistical analysis, and its use in SPSS. A typical problem involving weighting can be seen in the following example.


A survey has been carried out on the population of a certain region of the United States, to estimate voting preferences in the region for an upcoming national election. A sample of 1000 voters, including 500 African Americans (AA) and 500 Non African Americans (NAA) has been drawn at random, and the result was that 60% of the sample announced they would vote Democratic. The pollster in charge hastily made the prediction that Democrats would win by a near landslide margin. As it turned out, this was a wrong prediction: 55% of voters chose the Republican candidate, and just 45% cast the Democratic ballot.


What went wrong with the pollster's prediction is that it failed to take into account the fact that blacks were over-represented in the sample, and at the same time they were very strongly Democratic (about 80%) while whites and other non-blacks were under-sampled and they were also more evenly divided between the two parties. The prediction of 60% Democratic was based,
in fact, on 40% of the NAA sample and 80% of the AA intending to vote for the Democratic Party. As the same number of people from each ethnic group was interviewed, the overall simple average was 60% Democratic. But in fact, blacks were not even close to represent 50% of the population. Their share in the region was about 18% of the population. So in fact giving each portion of the sample its proper demographic weight (82% non-blacks, 18% blacks) gives a very different prediction