You can use the Weight component to define weighting, which you can then use to weight your data. Weighting is sometimes referred to as sample balancing. Typically, you use it when you want your data to reflect the proportions of various groups in your target population more accurately than your data actually does.

The Weight component is designed to work with the UNICOM Intelligence Data Model and requires that the CDSC that you are using is write-enabled (Can Add is True for the CDSC), supports the flat (VDATA) view of the data, and changing the data in existing records (Can Update is True for the CDSC). For more information, see Supported features of the CDSCs. In UNICOM Intelligence Professional, the Weight component always uses the flattened (VDATA) view of the data.

Sometimes weighting is used to treat the respondents as representatives of the total population of which they are a sample. For example, the data you collect in a survey reflects the attitudes of the people interviewed. However, sometimes you might want to weight the data to reflect the attitudes of the total population instead, as if you had interviewed everyone rather than just a sample of the population. This, of course, assumes that the people interviewed are a truly representative sample.

Suppose you interview 380 respondents from a population of 10,000 internet users, and discover that 57 members of this sample go fishing regularly. However, you know from national statistics that out of 380 people, normally 60 would go fishing. So you want to create crosstabulations that show the total number of internet users who go fishing as 60 instead of 57. Moving from 57 to 60 is the fine art of weighting. In this case, each internet user has an individual weight (sometimes referred to as the weighting factor) of 60/57.

Weighting can also be used to correct bias that builds up during a survey. For example, when conducting interviews by telephone you might find that 60% of the respondents are women. Using weighting you could adjust the results so that they more accurately reflect the more even balance of men and women in the target population.

The basic idea behind weighting is that the data for each respondent is multiplied by that individual's weight. For example, consider a simple table of age by gender. When the table is unweighted, the count in each cell is incremented by 1 for each respondent that satisfies the conditions of the cell. When the table is weighted, the count in each cell is incremented by 1 multiplied by the individual's weight.

The Weight component provides three methods of calculating the weighting:

Use this method when the weighting factors are already known for the respondents in each population group on which you are basing your weighting; for example, because they have been calculated using another software package.

See Factor weighting.

Use this method when you know the number or proportion of respondents in each of the population groups on which you are basing your weighting.

See Target weighting

Use this method when:

See Rim weighting

For all three methods, you specify the weighting parameters using one or more single response categorical variables. (You cannot use multiple response categorical variables to specify the weighting parameters.) Generally, you use categorical variables that record demographic information, such as gender, marital status, age, or region. However, the Weight component does not restrict you in the selection of the variables to be used. It is up to you to ensure that the variables that you choose are suitable. If necessary, you can create one or more derived variables to use to specify the weighting parameters.

The examples in this section illustrate only how to use the Weight component. They are not meant to be an accurate representation of how you should use weighting in a survey.