0
$\begingroup$

I apologize in advance... there's a good amount of stuff to read before you get to my question. Maybe I could have saved a word or two here and there... but overall, the background (I think) was required in order to best illustrate the topic.

Ok, I recently purchased SPSS in support of an upcoming research analysis. I've not used SPSS before and would like to get some guidance/pointers for normalization.

BACKGROUND:

As part of this academic research, I plan to release a survey in the next few weeks.

For my dependent variable (and dependent variable only), I will first provide the survey respondents a 2-column matrix where the first column lists the "name of a process" (across n number of rows) and the second column includes a checkbox.

Let's say I have 10 default rows (i.e., processes) from which the participants check any processes (in column 2) in which they believe their team contributes to.

Example: Survey participant #1 may check rows/processes: 1, 3, 7 Survey participant #2 may check rows/processes: 1, 4 Survey participant #3 may check rows/processes: 5, 7, 8, 9, 10 and so forth...

Using branching logic in the electronic survey, their selected processes (e.g., 1, 3, 7) will be passed into subsequent questions. Their themes are as follows:

Question #1 will require to indicate a frequency for how often a process may have changed. A survey participant will have to select a value -- on a scale of 1 through 7 -- for each process.

Question #2, #3, and #4 will ask something else (note: for the purpose of this blog question, it's not important what the actual questions read). For the 2nd through 4th question, survey participants will have to provide answers of either "Yes" or "No". (... where 'yes' will be coded as "1"; and 'no' will be coded as "0" in SPSS).

So far so good?

** Break **

Upon completion of the survey (I'm striving for several hundred participants), the data will be uploaded in SPSS. Now, again, all questions above are pertaining to a single dependent variable.

Based on the survey participants' answers, I'd like to create the following:

a) Individual score b) Overall organizational score

Obviously, multiplication across the columns will not work... as soon as, e.g., as a participant selected "no" in either the 2nd, 3rd, and/or 4th question, that process would be scored as zero. I don't want that! So, addition is probably the better choice.

Example of how survey participant #1 may have scored:

1: | 5 | 1 | 0 | 1 | = 7
3: | 2 | 1 | 1 | 0 | = 4
7: | 1 | 0 | 0 | 1 | = 2

Thus, participant #1's score = 13 (7 + 4 + 2)

On the other hand, survey participant #3 may have scored:

5: | 3 | 1 | 0 | 0 | = 4
7: | 4 | 0 | 1 | 0 | = 5
8: | 2 | 0 | 0 | 0 | = 2
9: | 7 | 1 | 1 | 1 | = 10
10: | 3 | 1 | 0 | 1 | = 5

Thus, participant #3's score = 26 (4, 5, 2, 10, 5)

Now, participant's #3 example score of 26 is twice as high as #1's example score of 13. This is primarily due to the larger number of projects/processes that the third survey participant is involved in.

As I'm trying to measure some form of "health" of the organization, a larger total score (derived through a larger number of processes) could be viewed as a contributing factor to the "organization's health". In this example, this would be wrong!

In other words, only larger scores on the individual process-level [(1 vs. 3 vs. 7) or (5, 7, 8, 9, 10) may be indicative of an "unhealthy" process/organization. So, process #9 by itself (with a score of 10) might be a contributor.

So, my question is as follows: In SPSS, how can I normalize the data so that it won't make a difference as to how many processes a survey participant has selected (and thereby skewing the data)? Should I merely divide by the number of processes? So, 13/3 = 4.333 vs. 26/5 = 5.2

If so, how can I accomplish that in SPSS?

I hope the above makes sense.

Thank you, EEH

$\endgroup$
  • 1
    $\begingroup$ If you are a bachelor student, simply taking the mean seems logical. $\endgroup$ – PascalVKooten Mar 19 '13 at 16:26
  • $\begingroup$ How is that done in SPSS? $\endgroup$ – user22000 Mar 19 '13 at 16:40
  • $\begingroup$ Open a syntax file and run the following code COMPUTE meanscore = mean(var1, var2, var3, etc...). EXECUTE. Oh and the reference: www.google.com $\endgroup$ – PascalVKooten Mar 19 '13 at 16:44
  • $\begingroup$ Var 1 = Question 1 Var 2 = Question 2 Var 3 = Question 3 ********** Based the description, am I not computing across Var 1 which as up to 5 scores (processes). Aren't we mixing apples and oranges? So, how will the individual selected processes across Var1 be stored for calculation? $\endgroup$ – user22000 Mar 19 '13 at 16:56
  • $\begingroup$ You could first standardize all questions, and then take the mean. I'm sorry, it is really difficult to read such a long question. I think when you click on menu Analysis, Descriptive statistics, you click on descriptives and there you can tick the box "save standardized variables". Select all your questions and then click on ok. You'll have many more variables, and then you can use that COMPUTE mean mentioned above. This will ensure every person has comparable scores, and the ones with higher scores having higher scores for real. (I ignore now that this could be called missing data problem) $\endgroup$ – PascalVKooten Mar 19 '13 at 17:14
1
$\begingroup$

Make new syntax file.

DESC VARIABLES= question1 question2 question3 etc /save.

This saved all the questions as new variables with a Z attached.

The following computes a single scale that you can use for analysis:

COMPUTE meanscore = mean(Zquestion1, Zquestion2, Zquestion3 etc)

$\endgroup$
  • $\begingroup$ For convenience: use CTRL + R to run selected syntax. $\endgroup$ – PascalVKooten Mar 19 '13 at 17:19

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.