i saw two explanation as shown in the following:
Negatively worded items are included in questionnaires to disrupt a response set where subjects respond favourably or unfavourably to all items (Marsh, Barnes, Cairnes & Tidman, 1984).
On the other hand, some claimed that the negative construct is excluded from the study. The reason is the expected uncorrelated relationship among negative construct with the other four dimensions has violated the assumption of a reflective measurement model as applied in this study.
Can the questionnaire contained negative construct?
The idea of reversing some of the items so that they are negatively worded is rather out of date. The problem is that some respondents will still use a "response set" and reply to all of the items as if they were positively worded. This will not only ruin your data for those respondents but also lower your correlations overall.
The alternative to negative wording is now sometimes referred to as "transparent" wording, so that if someone knows their feelings about the topic, they can work through your items as quickly as possible.
I would agree with first response, but I am not qualified to answer, just a guess
I haven't heard the term negative construct used for this issue before, but it is fairly common to include questions in a survey as a "check" to ensure people are responding to the survey honestly. For example, if a person responds "yes" to every question, but one of the questions is worded negatively, that would be a "check" that the person was paying attention. Usually these questions are some version of another question in the survey. In most cases, this question is not used in the factor analysis or scoring mechanism.
The idea of reversing some of the items so that they are negatively worded is rather out of date. The problem is that some respondents will still use a "response set" and reply to all of the items as if they were positively worded. This will not only ruin your data for those respondents but also lower your correlations overall.
The alternative to negative wording is now sometimes referred to as "transparent" wording, so that if someone knows their feelings about the topic, they can work through your items as quickly as possible.
Dear Huang,
Of course we can use negative direction or worded, reverse questions in questionnaire. This approach provides us with not only more reliability in terms of measurement but also ability to somehow assess the sincerity of our participant. But before we do analysis exploring the associations and causalities among variables, those reverse structured questions/items should be transformed accordingly. Regards.
Thank you very much to all of your guideline! The argument is the response bias cannot be addressed well by recoding. The problematic issue of negatively worded which will eventually be deleted in the final analysis. Recoding/ reversed item causes undesirable consequences: lower correlation of items, exhibit smaller factor loadings etc. So how to avoid the negative items?
I am not sure I understand what you're asking? We gave some answers to the question about negative responses, so what are you asking that's different now? Did we misunderstand your original question?
If you have a positive set of items (such as I like school, school is fun) and a negative set of items (such as I hate school, school is boring), such that the pairs of positive and negative items are in truly on a single continuum, you will get weak measurement trying to use an agreement scale on the items, because disagreeing with a positive item is NOT the same as agreeing with a negative one.
In this kind of situation, what you need is a semantic differential scale. In this kind of scale, you have one end of the continuum on one end and the other on the other end, and ask your participants to mark where on the scale they feel:
I hate school [_] [_] [_] [_] [_] [_] [_] I like school
Then you number their response from left to right. When I use these, I put 0 in the middle so I am coding -3, -2, -1, 0, 1, 2, 3. When I have a number of these items, a positive mean tells me that the responses average toward the positive sentiment and a negative mean tells me that the responses average toward the negative one. You can reverse the anchors so they vary the pattern, but the coding has to follow where the anchors are.
Dear Huang,
Yes, you can use reversed questions in your questionnaire. However, don't forget to recode them in the stats tool you are using (I'm using the SPSS).
Here is an example:
Q1: I feel depressive the last 10 days
(not at all) 1 2 3 4 5 (a lot)
Q2: I feel happy the last 10 days
(not at all) 1 2 3 4 5 (a lot)
The second question should be reversed so, you can recode it by creating a new variable in which 1=5 , 2=4, 3=3, 4=2, 5=1.
However, you will find that the "reversed" questions do not load with the other questions even when you reverse code them (check it with factor analysis). Disagreeing with feeling happy is not the same thing as agreeing with feeling depressed. If you have reversed questions and use a simple agreement scale, factor analysis will separate them into two distinct constructs every time.
So if I understand you correct Julia, we should analyze negative and positive construct questions separately, rather than recode them?
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Applied Statistics
- Survey Design