Using Case Weights

Using case weights is a complex issue, so let it suffice here to learn that, and how, it can be done with SPSS. All you need is a variable containing case weights (this variable indeed may be called "weight", but it may also have any other name). Then weights can be turned on with the command, e.g.,

WEIGHT BY weight.

and turned off with the command

WEIGHT OFF.

Note that this command should particularly used with "frequency weights", also called "replication weights". This means that the weights simply indicate that a certain constellation of variables appears n times in your data, with n being the weight given to a single data line representing this constellation.

Here's a simple example: Suppose you have a data set with five cases that have the properties "white" and "boy", six cases with the properties "white" and "girl", ten cases with the properties "black" and "boy" and finally eight cases that are "black" and "girl" (of coure, there are two variables involved: colour and sex). You may enter five rows of data with "white" and "boy" (possibly represented by any numbers you may choose), six rows of data with "white" and "girl", and so on. But you may also enter a single row with "white" and "boy" plus a weight variable with the value of 5; next there may be one line to represent the six white girls together with the value of 6 in the weighting variable, and so on.

However, if you have sampling weights that account for over- or undersampling, or perhaps even for different participation rates in a survey, you should have a look at the COMPLEX SURVEY commands offered by recent versions of SPSS. As yet, these are not covered by this guide; look for commands that start with CS in the SPSS help system or elsewhere.