I formatted a survey in SurveyMonkey using questions from a valid and reliable instrument (Principles of Adult Learning, Conti, 1978). Questions used a 6 point Likert scale (Always, Almost Always, Often, Seldom, Almost Never, Never/NA). When I formatted the questions in SurveyMonkey, I assigned values per the creator's instructions, e.g. for positive questions, Always = 5; for negative questions, Always = 0, etc. Respondents only saw words (Always, Almost Always, etc), not numbers. When downloading from SurveyMonkey to SPSS, is it necessary to recode the negative questions?
Thanks for your input.
When articulating the score weights assigned to survey item response labels (Sentiment, Frequency, etc.) it is critically important to have alignment of net polarity (higher vs lower) correctly assigned. If all of the items on the survey have expected responses that are aligned in the same net polar direction, then the distribution of the summed survey score for those items will accurately reflect the net polarity of the respondents for that cluster of survey items. If one or more of the items in the cluster of items being summed, is "reverse coded" ("Yes" means "No"; "No" means "Yes" etc.), and not corrected for the reversal, then the net summed score will never be valid because of responses can "cancel each other out" algebraically when summed. The short answer is follow the instructions for score weights that are assigned by the survey designer, or manually code and weight them yourself using common sense, in light of the logical response being solicited by the survey question (item). My bias is toward calibrating any and all weighted scales at a true zero, so that it is possible for a respondent to mark every item on the survey with the most extreme negative response (Strongly Disagree) and have a valid summed score of 0 for that respondents survey. Likewise, the maximum possible score will be determined by the 0 to Max range of score weights and number of survey items being summatively scored.
Example Sentiment (Agree-Disagree) survey:
https://tthsclubbock.co1.qualtrics.com/jfe/form/SV_0NDkNyjRUNbtdCR
Strongly Disagree = 0
Disagree = 1
Neutral = 2
Agree = 3
Strongly Agree = 4
Hello Victoria,
It sounds as if you have already taken into account the orientation of the questions in your (internal) assignment of values, so no further adjustment for polarity differences should be necessary.
The one feature that is potentially troublesome, though, is the "Never/NA" response option. To the extent that some folks interpret this as "Never," whereas others interpret it as "Not applicable to them, regardless of frequency," the option could result in being chosen for competing (and not-to-be-conflated) reasons. Did you try this out first, such that you are confident in it being interpreted unambiguously?
Good luck with your work.
Dear Victoria, in my oponion, a Likert scale is a ranking mrasure that has unequal intervals like an inteterval scale. But a response of N/A would be treated as missing data and therefore excluded from the analysis.
It is impossible to rank what does not exist. You may wish to skip your N/A option and repeat the analysis
From my understanding, you have addressed the needful as regard reversed items, thus there is no need to make any changes. The extracted data will be in a reverse form for the items reversely scored. All you need to do is compute the values and proceed with analysis.
I quite agree with some of the comments on the stand of not applicable/ never. However, there are some situations that the item or question do not apply to the respondents among others. making a tick on never might mean that you have been in such situation before, but never experience the impact, while NA (not applicable) means that you have never found yourself in such situation.
There are extreme cases where this could be found in questionnaire, however, it is best to avoid such if possible.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS