I have collected data through survey from in 5 point likert scale where 1 as Strongly Disagree to 5 as Strongly Agree. I have total 20 questions in 4 constructs: economy, psychology, political and social. Among 20 questions, around 8 questions are in negative statements. So, after entering all data into SPSS file; I did reverse coding for those 8 statements of negative meaning. Here, I am in confusion for the next step; that should I change these 8 negative meaning statements into positive meaning statements? If I have to change, what should be the format of sentence?
For example,
negative statement: I am a bad student.
positive statement : I am not a bad student. (option-1)
or, I am a good student. (option-2)
If I have to change the sentence, which option will be appropriate?
Thanks and regards.
No – leave the questions and answers alone. Just map the answers to numeric values in the reverse order, so strongly agree is mapped to 1 rather than 5.
You should note that negatively worded questions do not behave like positively worded questions in reverse. The first problem is that a negative question can produce double-negative answers.
Q: I am not worried about climate change
– the answer " somewhat disagree" means "I am somewhat not not worried about climate change".
The second thing is that even if the question is not a negative statement like the example above, the question itself produces a cognitive set – an interpretation framework – that is different for positive and negative thoughts, experiences, attitudes etc. One famous study asked participants to pay attention to their breathing. Every half hour they rated their breathing. Half were asked to estimate how much time they had been able to breathe freely while the other half were asked how much of the time their nose felt stuffed or blocked. Those who were cued to attend to breathing freely reported more time breathing freely than those who were cued to attend to the opposite.
There are some scales with negative items in which the negative and positive items actually form separate factors!
Once you have reverse coded the items, you don't have to do anything with the statements. Re-wording the items could affect its validity. Therefore, just compute the total score of each domain and report it. I have a Youtube video on reverse coding that might be useful.
https://www.youtube.com/watch?v=u5BO5vNOIP4
The next question is what to do with them statistically. Often the reverse scored items create a second factor highly correlated with the first (try a CFA with the positive and negative items loading on separate correlated factors). The question is then is this an artefact (e.g., often reverse scored items are grammatically more complex and have slower response times) or are the negative not example opposites of the positives. In your example, are bad and good students not exact oppositive (in your example, does a bad student not do homework, performs less than expectations [but may still do okay] and mis-behaves, while a good student gets good grades [and the opposite of that might be a student who works hard but fails).
Interesting paper here on the CES-D whose negative items show a response bias
1.Siddaway, A. P., Wood, A. M. & Taylor, P. J. The Center for Epidemiologic Studies-Depression (CES-D) scale measures a continuum from well-being to depression: Testing two key predictions of positive clinical psychology. J. Affect. Disord. 213, 180–186 (2017).
Daniel Wright If the reversed items do correlate in the way that you suggest, that should be obvious in an EFA with correlated factors, without taking the extra step of getting into CFA.
David L Morgan , I am not sure I understand your comment. Are you saying doing an EFA first even though you believe there might be this aspect to take into factor, and then only if you observe something "obvious" in the data "testing" this with these same data in a CFA (i.e., confirming what you just saw)? I assume this can't be what you meant, so hopefully you can help me understand what you meant.
Yes, if you think there might be a "wording" factor, then do an EFA with correlated factors first to see if that is a reasonable hypotheses. My suggestion was to do this in order to save the time and effort involved in using CFA for such a simple question.
Okay, I guess I did read that correctly. My fault. I hadn't thought that anyone would suggest that the seconds of time saved were so important nor that if you observe an "obvious" pattern then that pattern becomes a "reasonable hypothesis" that you can test with the same data nor that procedures like CFA are not useful unless the result was obvious before you used CFA (in these examples the factors are usually highly correlated so in many cases things like a scree will suggest one factor). One nice thing about ResearchGate is that you get to hear other people's thoughts and learn about different approaches from different research traditions! I have learned!
- Badges
- Science topic
- Similar topics
- Biological Science
- Microbiology
- Microorganisms
- Bacteria