Creating Composite Scores in SPSS


Composite scores

  • One of the main flexibilities with a statistical package like SPSS (besides being able to conduct statistical analyses and obtain tabular and graphical output) is the capacity for manipulating data.
  • One important data manipulation is the creation of new variables, which are some mathematical function of other variables.
  • Psychological research often involves measuring fuzzy constructs, such as personality traits, by gathering responses to multiple items (such as questions in a survey) which are combined to provide a (hopefully) reliable and valid measure of a broader construct (such as extraversion).
  • These "composite scores" can be:
    • Unit-weighted, with the data from each item being equally weighted (by either adding all items together or calculating the average of each item), or
    • Regression-weighted, e.g., from a factor analysis (not explained here).

Via pull-down menus

  • Make sure you have the data file open, then go to Data View (.sav)
  • Enter new variable name in "Target Variable"
  • Enter formula for creating composite score in "Numeric Expression"

  • To compute the score, click OK.

  • Scroll to the right-hand side of the data file (in Data View) and you should see a new column. 

  • If the the new variable doesn't have any data, try saving your data file (which executes any pending calculations).

  • If data appears, then its a good idea to run frequencies, descriptives, and/or an appropriate graph to check whether you have the kind of data you intended to create.

Via syntax

  • An easy way to create the syntax commands is to follow the instructions above, but on the last step, click Paste instead of OK.

  • The syntax can then be conveniently copied and edited - this is especially useful if you have many composite scores to calculate.

  • It also allows the syntax commands to be saved (.sps) and recalled for later use.

  • e.g.,

compute satistotl = satis+satis2+satis3+satis4+satis5.

compute satismean = mean(satis1,satis2,satis3,satis4,satis5).

Allow for missing values

  • You can allow for missing values by adding ".X" after "mean", where X is the minimum number of variables that need to have data for a case in order to calculate a mean.

  • For example, this syntax will calculate a satis score for a case as long as it has at least 3 values of the variables listed:

compute satismean = mean.3(satis1,satis2,satis3,satis4,satis5).

  • If a case has data for less than three of the variables, satis will be system missing (.).

  • Otherwise, a mean will be created using data from all available variables.

See also

  • Francis, Section 4.5 "Creating New Variables"