Hello Friends,
My PhD thesis is a cross sectional study which will be using survey method. A part of my research require collecting information about both dependent and independent variable from the same participant.
I have been suggested by some that in such a case, it would be a better idea if the survey is designed in such a way that the information on dependent variable is collected first and on independent variable should be collected later.
I want to know
1. Is this method fine or should I go for independent variable first and then dependent variable
2. Is there a paper which I can read and even cite regarding this issue
Thanks in advance
There are two main shortcomings for a cross-sectional design using the same source.
1. common method variance: the correlations among variables can be inflated or biased because the same person assess all the variables. Changing the order of IV and DV cannot solve this issue. To reduce this bias, read the following article.
Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., and Podsakoff, N. P. (2003), ‘Common method biases in behavioral research: A critical review of the literature and recommended remedies,’ Journal of Applied Psychology, 88, 879-903.
2. causality: we cannot test the causality with a cross-sectional design study. If you assess DV first, this will cause more problems in terms of causality.
I strongly suggest you to use a time-lagged design (e.g., assessing IV first, after at least two weeks or one month assessing DV.).
Thanks a lot Tae-Yeol Kim for the information. I had gone through the paper Podsakoff et al., 2003 and I do understand the issue of common source bias and the issue it can raise regarding causality.but as the current data stands, collecting data at two times is not a possibility so the only option I had was to look at minimize the bias.
In constructing a questionnaire it is highly recommended to put the sensitive questions at the end of the instrument such as asking personal and demographic charactristics
But if you mean by independent variables other variables such as past behavior or habbits or activity or life style then it does not make difference if they are asked before or after the dependent variable
Insofar as the common methods / single respondent issue, the point is that respondents may "guess" the relationships being studied. Placing the dependent variable early in the survey helps reduce this by if is not obvious which it is.
Regarding common methods in general, I would encourage you to look well beyond the Harman test, which has poor psychometric properties and is increasingly considered insufficient evidence. For an overview of solutions, some of which you can implement during study design, see Appendix C and text of:
Krishnan, R., Martin, X., & Noorderhaven, N. G. 2006. When does trust matter to alliance performance? Academy of Management Journal, 49(5): 894-917. (Available at http://scholar.google.com/scholar?oi=bibs&hl=en&cluster=3292640590399664878 or via request.)
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Applied Statistics
- Survey Design