Is a variable always in numerical form Recoding with SPSS

At the same time, SPSS requires numeric data for many functions, where data is in String (or text) format. In the case of pure text data, such as 'Very good' or 'Good', SPSS cannot recognize that 'Very Good' reflects a higher intensity or quality than 'Good'. However, SPSS would recognize a meaningful ranking in numerical form.

SPSS allows String variable automatically in numerical to encode.

To do this, use the TRANSFORM - RECODE function in the menu bar. Now you can decide whether you want to recode to the same or a different variable. It is better to stand up for another variable to decide, because the recoding (you can also recode several values ​​to a single new one) can lead to a loss of information (whether deliberate or due to an operator error). This problem is eliminated by recoding to a new variable.

You now select the variable that is to be recoded and enter a new name for it in the Output variable field (which for reasons of various compatibility) must not exceed eight characters. In the field below, however, you can choose any long and expressive name.

Figure: Recoding into other variables with SPSS

Now click on Old and new values. If you want to recode individual values, enter in the field value the same one (e.g. Matura). Now you need a new value for it. For this you could e.g. the number 3 in the field New value enter. In the example given above, various completed training levels would be put in a logical order.

If you want to combine several (numerical) values ​​into a single new one, you can specify a range (e.g. Area 20 to 29), if you want to bring everyone between 20-29 year olds into a single age group ’between 20 and 30’). Click after each individual entry for transcoding Add.

At the end, click Next and then OK. Your data will now be recoded into the new variable. Only now would it be possible for you that SPSS can meaningfully rank the training according to years and quality and, of course, can also calculate many other (related) analyzes. Only now would it be possible to calculate a correlation between the level of education and income.