I am conducting research on the effects of residential environments on mental, physical, and social health. In my questionnaire, mental, physical, and social health are measured by three separate scales, and some of the questions in the residential environment are also measured by the scales. So, when I conduct the validity test (exploratory factor analysis), should I test the validity of these four scales together. Or should I divide them into Mental Health, Physical Health, Social Health and Residential Environment and conduct four separate validity tests?
Also, my questions on residential environment are actually divided into three dimensions: indoor, outdoor, and physical environment. However, each question actually corresponds to one independent variable, rather than synthesizing one independent variable through multiple questions, and the questions within these three dimensions are not actually very correlated. So instead of analyzing the three dimensions separately, can I just analyze the entire living environment for reliability?
First of all, and seriously, you may be trying to do TOO MUCH in one survey! Looks nice for the researcher, but is often tedious to the interviewee. That leads to a lower quality of answers as the survey progresses, and detracts from a willingness to participate in future surveys.
Second, "validity" is about the objectivity and comprehension of the survey questions and answers. The type of ranking/scales used is irrelevant - unless they introduce a bias to the answers.
Third, you don't "synthesize" an independent variable. Something either is an independent variable, a dependent one, or irrelevant. An independent variable may affect different dependent variables in different ways. Your questions may discover that. But again, it is MUCH better to keep a survey to less than 20 questions, and return later for another short survey, than to try to load one survey with everything that you think you want to consider, and then get bogged down with poor hypothetical questions.
Ningbo Kou You can first perform factor analysis, then converging and discriminant validity. Then conduct a reliability test of each scale within the questionnaire. This could be done through Cronbach's alpha (consistency test) and then test-retest (reliability test).
- Badges
- Science topic
- Similar topics
- Indexes