Hi everyone.
I am investigating how psychological factors influence human attitudes and behaviors towards environment.
For that, I have an SPSS dataset of 1076 respondents (49% response rate) that built from responses on a questionnaire of 6 themes (3 independent variable & 3 dependent variables). There is average of 13 questions for each latent construct. Each question measured on 7-point Likert scales (-3 to 3, with 0 as neutral point).
However, I noticed that some respondents give repetitive answer (e.g., answer is 0 for all latent construct). I tend to not trust this data, as it might be the respondent did not fill it seriously.
How can I deal with this issue? Should I keep it, or delete it manually, or is there any specific instruction in SPSS that can help me to delete it automatically?
Thank you.
Dear Zulkhairi,
I think you are wise to question the validity of such responses. I would suggest that it is easier to identify patterns and deal with them collectively in a spreadsheet package, such as Excel, such as using a sort option and separating the affected records. You can easily download data from SPSS into Excel and reload it back again.
Personally, I always do pre-processing in Excel before uploading data into SPSS.
An alternative might be to replace these 0s with null data as the rest of the records might still be usable and SPSS handles missing data very well. Again, identifying the records with groups of 0s is easier in a spreadsheet package.
Dear Zulkhairi,
first of all, your questionnaire should be validated by at least five experts who are specialist on your scope and era. after face validity, you should do pilot test to edit questions for better understanding and reliability. for those cases, which respondents answer repeatably, they are outliers and should be omitted. if your model is complex, you must analyze it using SEM (structural Equation Modeling) .
Dear all. I really appreciate your response. It took me some time to get the result of MD but I found this analysis very interesting.
I would like to get clarification on other things. In my case, three of my variables were indices developed by computing a mean from 13-15 items. Then the average of every mean from the first three variables is used to create another variable. I have performed MD on these 4 variables, and make a list of every problem cases. From here, I compare this list with another table that listed outliers & problem cases of same variables, identified using boxplot and Explore option. I found very few problem cases overlap in these two tables. For example, MD identified my 100th case as problematic, however, it is not categorized as outliers/extreme in boxplot/Explore option for the same variable. How can I make sense from this comparison?
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS