0
$\begingroup$

I have three variables measuring different aspects of performance each with a scale of 8 and I want to create a new Performance variable by computing these together. I have done this the following way: I've gone to "compute variable" in SPSS and chosen (var1+var2+var3)/3. Missing data is not a problem so I didn't have to add anything or that in the formula. Everything seems to be fine until this point and I only get full numbers between 1 and 8 by choosing "no decimals".

However, when I take the descriptive statistics for this new variable, it gives me the same number multiple times. Because of this, I am not sure if I have created the new variable in the right way and if I can start running regression analyses

$\endgroup$
2
  • 1
    $\begingroup$ (1) Not sure what do you mean by 'it gives me the same number multiple times'. If you have only 3 variables with only 8 possible values AND you are rounding to integers I would not be surprised if a lot of respondents had the same average. (2) Just as a suggestion you could also use the MEAN function instead of adding and dividing. See for instance kb.iu.edu/data/adjn.html $\endgroup$
    – David
    Jul 16 '12 at 19:36
  • 1
    $\begingroup$ Any reason for computing a mean score instead of a sum score? $\endgroup$
    – chl
    Jul 17 '12 at 10:54
2
$\begingroup$

This is because in the background, a number like 5 can actually be 4.67, 5.00, or 5.33. However, since you have suppressed decimal place to 0, the output table will appear to have 3 rows of 5s.

Suppressing the decimal place at the variable tab is only decorative, in order to truly get rid of all decimal places, you'll need to work on it at the compute stage.

This is your original score:

compute myout1 = mean(v1,v2,v3) .
execute.

If you'd like to truncate:

compute myout2 = trunc(mean(v1,v2,v3)) .
execute.

If you'd like to round up:

compute myout3 = rnd(mean(v1,v2,v3),1) .
execute.

If you run a frequency on myout2 and myout3, the multiple lines should be gone.

On a side note, averaging items and then using their mean as a variable can sometimes be too rough because the underlying assumption of the three items carrying equal weights is often challenged. If you have a large enough sample size, consider modeling them differently, such as putting them into the model independently.

$\endgroup$
1
$\begingroup$

Try using the keyword TRUNC, e.g. TRUNC(MEAN(var1,var2,var3)).

$\endgroup$
1
$\begingroup$

You are not going to get just whole numbers this way. Displaying values with no decimals is just a formatting choice. The fractional values arising from your denominator will still be present.

If you just sum the three variables, you are certainly likely to get some results that are the same for multiple cases. Try a histogram of your new variable to see what the distribution looks like (Graphs>Chart Builder).

But thinking ahead, what are you planning to do with this variable? Why do you need to boil it down to one?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.