Dear Researchers,
I am doing my phd thesis in second year and need to consult you about survey scale, my survey will be online and I am not sure which scale helps prevent bias responses: bipolar (2statments) VS unipolar (one statement)?
Some research saying that bipolar reduce the bias and helps for more reliability and validity, but I have some concerns about how participants respond on two statements easily without difficulty, also how creating bipolar survey online! Unipolar 7-point scale with one statements is still good, very common and can be trusted as well!
Actually it is a baffling matter!
Can you help me with that please?
Thanks and Best Regards,
Kemo
Items that ask about only a single rating factor (e.g., degree of agreement or satisfaction) are easier to interpret.
Hello Khalid,
I would guess that the specific construct/concept/object/sentiment that is the intended focus of the measure would potentially have something to do with the "best" choice for a given perception inventory. Both approaches have been used profitably in the past. The semantic differential (Osgood, Suci, & Tannenbaum, 1957, The measurement of meaning) is a well-known example of a measure that used bipolar response scales.
However, the semantic differential is different from a forced-choice (two-statement) approach, in that the forced-choice approach doesn't allow for a neutral response. There are instruments that use the forced choice format, such as the Myers-Briggs Type Indicator.
I agree with David Morgan that, all other things equal, the simpler the stimulus, the less likely respondents will misinterpret or misunderstand the task.
If you're more concerned about whether specific, web-based survey applications support particular response formats, I suggest you look at those before choosing the application. Here are two starting sources:
1. https://help.surveymonkey.com/articles/en_US/kb/Available-question-types-and-formatting-options
2. https://www.qualtrics.com/support/survey-platform/survey-module/editing-questions/question-types-guide/question-types-overview/
Good luck with your work.
I have used unipolar scales and they worked without necessariy increasing the response bias. For example, I would prefer using scale of 1-5 with 1= totally disagree, 5= totally agree. The bipolar version is hard for the respondents. For example, a scale with a 0 in the middle and -5 at extreme left hand end (totally disagree), and +5 at the extreme right hand side (totally agree).
Dear Khalid,
I performed more than 20 surveys with the both scale. If you ask about the agreement with some more or less abstract facts (statements), unipolar scale is better, if you need an opinion on some real things, for example ("the quality of our product is better, the same, worse than that of or direct competitors") a bipolar scale is quite comfortable to respondents. In addition, it is easier to find systemic biases in answers. In one of my surveys of 1000+ CEOs of manufacturing companies the assessment of costs and prices followed the normal distribution with the mean and median around zero, the assessment of quality was largely biased towards mean and median +2 ("much better than that of competitors").
- Badges
- Science topic
- Similar topics
- Statistics
- Survey