like in stata how one may replace for missin values ussually given by a (.) into something like 0
First, click on "Transform", then "Replace Missing Values..." in the toolbar at the top of SPSS
If you have decimal points (e.g observations to 1 or 2 d.p.) this can create problems - if not replace any 1 or 2 d.p. Observations with whole numbers, then use transform - replace missing values
RECODE VarList ($SYSMIS = 0).
EXECUTE.
Replace VarList with a list of variables from your data file.
To explore the subject more deeply , I recommend you to study the methods for "imputation of missing values" in Tabachnick & Fidell: Using Multivariate Statistics. If you can, see the little book Allison, Paul: Missing Data (Sage). I linked to a brief exposition of the latter.
http://www.statisticalhorizons.com/wp-content/uploads/2012/01/Milsap-Allison.pdf
How can I replace missing values in SPSS with the mean of a number of items?
Hello Forough. Please provide the context for your question. Why do you want to impute means? What is the ultimate goal, and what analysis will be performed afterwards?
Thanks Bruce for your attention to my question. I am going to run a CFA on a 15-item fatigue scale. For Items 4-14, respondents are allowed to check the "not applicable" box. if a respondent checks the "not applicable" box for an item, no score is assigned to this items. They are different from missing values.To calculate a global score, mean of items 4-14 is calculated. To run a CFA, items with no score must be replaced with the mean of items 4-14. Manually I can replace them but it is difficult.
Suppose your original variables are called F1 to F15 (F for fatigue). And suppose 88 is the value you used to indicate N/A for variables F4 to F14. I would start by making copies of F4 to F14 so as to not overwrite the originals (which may be needed for something later).
DO REPEAT old = F4 to F14 / new = Fcopy4 to Fcopy14.
- COMPUTE new = old.
END REPEAT.
Next, I would make sure that whatever value indicates N/A is treated as missing. Suppose that value is 88; and suppose that 99 is used to code missing values.
MISSING VALUES Fcopy4 to Fcopy14 (88 99).
Finally, replace missing values with the mean of valid values for Fcopy4 to Fcopy14. I'll assume you want a minimum of 5 valid values in order to compute a mean. I plucked that number out of thin air. Feel free to change it to a more appropriate value.
DO REPEAT x = Fcopy4 to Fcopy14.
- IF VALUE(x) EQ 88 x = MEAN.5(Fcopy4 to Fcopy14).
END REPEAT.
EXECUTE.
NOTE that this is untested. But it should give you the general idea. And of course, you'll have to replace my variable names with your own.
HTH.
Not sure this is exactly the challenge I faced. I had about 100 columns (variables), many with missing data. I needed them to be zero(0) instead of dot(.).
This is what I did:
1) I exported the data to excel (You can also copy from SPSS and paste in excel)
2) I selected the required columns.
3) I pressed Control + H (Find and Replace)
4) I replaced all blanks with zero(0)
5) I copied the data from Excel, and paste in SPSS.
That's it!
It took less than 2mins to do that. But it took me over 5hrs to think that simple thought.
Using the [Recode into Same Variable] function.
From Transform Menu --> Recode into Same Variable --> Old and New Variables --> System Missing --> in value space add the value you want to replace the missing data with --> continue --> Ok.
Done.
Good luck Roy Naburuki however i know for sure you already resolved the problem.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software