Specially, in mixed-word scale, respondents may variably rate items which are under the same construct (e.g., 2, 4, 5, 1). Is deleting the such cases acceptable in data cleaning stage?
You can do anything you like as long as you explain what you did. There are no rules :)
David L Morgan
1. I mean responders variably rate the items (Linkert scale) in the same construct. For example, a construct includes 4 items. Responders rate them 5, 3, 2, 1 respectively. If this case happen, can I delete the responder who scale like this from the sample group because as I know, the Cronbach reliability of the construct may be low. Is it acceptable in academic sight if I delete the responder.
2. Moreover, the existing theory proposed that a variable has multi-dimensions which are in positive relationship. For example FFMQ (five facet of mindfulness) comprises observing, describing, actawarness, nonjuding, and nonreacting. In contrast, my current study found that some participants response or rate the questionnaires of actawareness contrast to the other constructs which may lead to negative relationship between actawareness and other dimensions.
I presumed that this phenomena occur because of their confusing with the mixed-worded items in the scale (positive and negative worded items; items of actwareness are negative word, but others are positive). Based on this situation, is it acceptable if I delete the such participants before data analysis or I should accumulate them in data set and analyze. Whether the result of the relationship between actawareness and others is negative or not, I should report them.
Thank you
Just because items are correlated, does not mean that everyone will use the same pattern in responding. Rather than looking for patterns in responses, an alternative approach would be to form the scales you propose and then examine them for outliers.
Have you used to Confirmatory Factor Analysis to assess the relationships between your proposed scales?
David L Morgan
Yes sir, I used CFA to assess the measurement model. The scale can not meet the fit. Moreover, I conducted the alternative model. It met good fit, but the factors loading are low which effect to failure of CR & AVE. However, I will try to examine the outliers.
Thank you very much
- Badges
- Science topic
- Similar topics
- Statistics
- Survey