Witold, you could do this easily using the SPSS menue:

- Transform

- Recode into same variables (chose all 200 variables)

- Old and New values: Once you click Old and New Values, a new window where you can specify how to transform the values will appear. There, you can chose to transform 'system missing' into 999.

Regards, Karin

Witold, you could do this easily using the SPSS menue:

- Transform

- Recode into same variables (chose all 200 variables)

- Old and New values: Once you click Old and New Values, a new window where you can specify how to transform the values will appear. There, you can chose to transform 'system missing' into 999.

Regards, Karin

there is an online electronic stats book called stat-soft which has a good chapter on mean substitution or elimination case-wise or list-wise of missing data I think it is, I could be slightly wrong in my terminology

You can use keywords TO and ALL with RECODE.  If your 200 variables are contiguous in the file, you could do this:

RECODE var1 TO var200 (SYSMIS=999) (ELSE=COPY).

You can use multiple instances of TO if there are multiple sets of contiguous variables.  E.g.,

RECODE var1 TO var10 var100 TO var110 (SYSMIS=999) (ELSE=COPY).

 You could also try using the keyword ALL, like this:

RECODE ALL (SYSMIS=999) (ELSE=COPY).

I don't have SPSS on this machine, so cannot test right now whether ALL works in a file that contains both numeric and string variables.  Just give it a try, and see what happens.  HTH.

Dear all,

In terms os SPSS the easiest way is = Karin.

But doing what Bruce said you need to work with syntax of SPSS. Then the 2 ways are very good in my opinion, but it depends on what is the most easy way for you.

Have a nice day

Helena

I now have access to SPSS again, and have tested using RECODE with the keyword ALL in a file with both numeric and string variables.  It does not work.  Here is the error message:

>Error # 4636 in column 12. Text: (

>On the RECODE command, all the variables to be recoded must be of the same

>type. That is, all must be string variables, or all must be numeric

>variables.

>Execution of this command stops.

Hello Bruce, 

Have you automatic recode the string variable?

Helena

Hi Helena.  No, I was simply trying to find out whether RECODE with the keyword ALL would work in a file with a mixture of string and numeric variables.  The answer was NO, it doesn't work.  Here is the syntax I used to demonstrate.

* Create a small dataset with 3 numeric and 3 string variables.

DATA LIST list / n1 n2 n3 (3F1) s1 s2 s3 (3A1).

BEGIN DATA

1 2 3 "A" "B" "C"

. . . "" "" ""

END DATA.

* The following RECODE command will generate

* an error message, because all variables have

* to be of the same type (i.e., all numeric

* or all string).

RECODE ALL (SYSMIS=9).

LIST.

* The next two RECODE commands will work, because

* the variables are all numeric for the first one

* and all string for the second one.

RECODE n1 TO n3 (SYSMIS=9).

RECODE s1 TO s3 ("" = "9").

MISSING VALUES

  n1 TO n3 (9) /

  s1 TO s3 ("9").

LIST.

Copy your variable data into Excel.   Fill all blank cells  with a number such as 9. Copy the column variable after filling all blanks and paste in SPSS . Define 9 as missing in SPSS.

Helena wrote:  "In spss you recode missings for numeric variables. Then, if you have strings with missing, 1st you have to recode strings to numeric (automatic recode) and then because all the variables are treated as numeric , you can recode all."

Hi Helena.  Yes, given a file with some numeric and some string variables, you could use AUTORECODE to create new numeric versions of the string variables.  If you then deleted the original string variables (via DELETE VARIABLES command), only numeric variables would remain in the file, and you could use the keyword ALL in a RECODE command.  But note that there is no guarantee that blanks in the string variables would end up coded as 9 (or whatever your missing value code is) in the variables created by AUTORECODE.  So a bunch of additional RECODE commands might be needed first.  In short, I think it would be easier to have one RECODE command for numeric variables and another for string variables (using keyword TO, perhaps), as in the example I posted earlier. 

RECODE n1 TO n3 (SYSMIS=9).

RECODE s1 TO s3 ("" = "9").

MISSING VALUES

 n1 TO n3 (9) /

 s1 TO s3 ("9").

LIST.

Cheers,

Bruce

Sorry for asking this basic question:

How to run above commands? Where should I type them?

I'm new in using SPSS.

Thanks,

Mehdi

Mehdi, click on File > New > Syntax to open a new syntax window, and type (or copy and paste) the commands there.  Then click on Run on the top toolbar to see the various options for running the syntax (e.g., Run-ALL, Run-Selection, Run-to end).

And note that when you are using the GUI, you can generate command syntax by exiting the GUI via the PASTE button rather than OKAY.

HTH.

Dear Withold, 

Usually if in Variable View you identify the categories correspondent to Missings , in terms of data analysis you would not have any problem . 

But  if for example you want to aplly "StaFs" for open question analysis, then you need to replace all the missings . You can do it in another more "complicate" way like this:

1º) You must first of all with your mouse "sort ascending the column of the variable you want to replace the missings. 

2º) you go to "Transform Compute  " and in "target variable" you introduce a name "ex ID", and in "Numeric experession" you enter the function "$Casenum" and press "OK".

3ª) then in your data file you will have a column with the nº from  "lots of missings" first until it begans with the nº1 to n of your data .

4º Then you make a copy of your variable "ID" calling it ex "ID_R

5º Then, if "ID_R" is