As the name implies, recode means replace some or all codes of a variable by new, different codes.

offers you two choices

• Into Same Variables: recode a variable "in place", i.e. actually changing the codes of a variable and destroying the original coding.
• Into Different Variables: recodes a variable by creating a new variable, i.e. leaving the original variable untouched.

When in doubt, always create a new variable, as this leaves you with the original untouched, i.e. in case of a mistake (bad recode specifications) you can redo the recode again...

Each form of the command produces a slightly different dialog box:

Just select the variable or the variables you wish to recode.

Select the variable or the variables you wish to recode, and for each of them indicate an Output variable: give it a name and a label, indicating somehow that it is a derived variable. When you hit the Change button the ? mark in the first field will be replaced by the name of the new variable to create

Then click Old and New Values to get this dialog

and you will get this dialog if you recode a variable in-place (fewer possibilities on the Output side.

A recode is usually a sequence of various recode specifications instructing SPSS to change some old value into a new value. The dialog boxes show also a list of specifications already entered (added) using the Add button.

• Value: Specify a particular value to be recoded
• System Missing: All SYSMIS values will be recoded
• System or User Missing: All SYSMIS and all user missing values (i.e. missing values declared by the user without any need here to indicate the code.
• Range .. through ... : A range of values. Make sure to understand that SPSS variables might be decimal values i.e. a range 20 through 30, followed by a second range specification 31 through 40 misses all values between 30.00001 and 30.999999.
• Range lowest through ... : Any value below a limit.
• Range ... through highest: Any value above a limit
• All other values: All other values, i.e. values not handled by one of the specifications.

• Value: New value to assign.
• SYSMIS: Assign to SYSMIS

Several recode specifications have already been entered...

New value section (recode into different variables): Same specifications as above, plus

• Copy old values: Just copy the old value(s). Especially useful with ranges if you want some values left untouched.

In addition to these possibilities, lets you transform numeric variables into string variables or string variables into numeric variables. Check the documentation if you need to know more about this.

The first Recode dialog offers also an IF button that lets you perform a conditional RECODE, i.e. you only recode the cases corresponding to a logical condition. E.g. you could perform a particular recode for women only, but not for men.

Recode specifications list what old values you want to recode into new values; all values not specified will be left untouched, i.e. stay the same as they were before starting the recode, i.e. if you "forget" some values they will simply be left untouched. This is quite obvious with , but also for but with different consequences: if the target variable already exists (SPSS will warn you), only the cases matching the specifications will be changed (overwritten), all others will be left untouched. If the target variable does not exist (new variable), SPSS creates it and sets all values to SYSMIS, therefore if your recoding scheme does not cover all values of the original variable, the corresponding observation in the new variable will stay SYSMIS.

If your Recode creates a new variable, no properties will be defined except the variable label if you indicated it with the target variable name.

In all other situations, RECODE only affects the data values, i.e. value labels and other properties will remain as before, unless you change them explicitly. Assume a simple example of a sex of respondent variable coded 1 for male and 2 for female with appropriate labels. Now recode 1 into 2 and 2 into 1 (reversing the codes)... but the lables will stay as before, i.e. code 1 is still labelled "male", although it now corresponds to women... So please be careful to modify the variable properties immediately... That is for this reason that I recommend to recode into a new variable to avoid this kind of trap. []

• Autorecode: automatic recodes
• Visual Binning: Recode continuous variables visually.