I have come across a situation regarding the application of PLS SEM for my research scenario. I have survey data of 213 respondents in total with Likert scale responses ranging from 1 (Strongly disagree) to 7 (Strongly agree). The 213 respondents get to answer the same first 9 questions, and then the questionnaire splits into 2 sections based on the user's response of question number 9. Each respondent is then able to answer an equal number of questions from either of the sections. Now, if a respondent answers the questions from section A, his responses to questions from section B seem as 'missing'. This happens because all responses to each of the questions are stored in unique variables. Hence, these 'missing' values are not missing by the user's choice but rather by the construction of the questionnaire. How should I treat these 'missing' values? As 'Not Applicable' values, zero values or something else? Please guide me on how to process this dataset to be used immediately for SEM Analysis.
Any and all help is appreciated.
Dear Faizan,
The question isn't so much about the software as the reasoning. What is the relationship between the two sets of questions? Because for most reasoning we would not pool answers to different questions. If you think that there is a direct correspondence, then construct some new set of variables where you take whichever answers were given along with some record of which group a participant is in.
Cheers.
Since it is only about 200 cases, could you do a manual data cleaning with excell and load data you are sure of so that the software automation does not stand in the way?
Good luck
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling