In my study, I am using a standardized questionnaire for obtaining data on different factors. The questionnaire contains 98 questions. The first 9 questions are on 7 point likert scale ("0=Never" to "6=Always") and rest of the questions are on 5 point likert scale ("1=strongly disagree" to "5=strongly agree"). I wish to use a 5 point likert scale for all questions. What are the possible issues in such a conversion of scales?
I don't understand why you would want to do that. It is clear that the first 9 items ask to an intensity of an activity being performed while the rest are based in intensity of agreement. You clearly have two sets of items you do not want to be converting into one and the same scale. It must be two separate concepts. I do not see how you could have the same scoring procedure without the need of rephrasing the items. If that is possible at all.
If I am correct,you want to normalize scales to they can be used in further calculations. Save all your variables as z scores (SPSS - Analyze, Descriptives, ...?), Tick the box save z scores. Done. Now all your variables become from -1 to 1 and you can use them jointly.
Best,
Eugene
I send to him two files that are in Spanish.
He can translate them with any Spanish-English translator.
It might be realized.
I have put several examples to him of howrealizing it.
Greetings.
Rafa.
Thank you for the responses! I do realize that the scales used in the questionnaire would have been designed as per the questions asked, however, the difference in scales can be a minor issue while dealing with the data. Conversion to Z-scores is a great solution to it.
Maybe you might reconsider transformations to z-scores for all items. If effect what you want to have is have a transformation that get you from 6 t o5 codes with a linear transformation. That is impossible. Two of the 6 categories have to be combined to give you one category, but the situation is worse. The 5 categories probably will have a neutral mid category that the 6 categories does not have. The six will have three positive in three negative categories. You could recode the two least extremes in to the less extreme five category. If would be from 6 to 5: 0 =1 , 2,3 =2,4,5 =4 and 6 =5. In that way you will have no middle values as you do not have these in the 6 category case. If you create z-scores and your sample is large and the distributions are skewed you could get very large negative of postive z-scores. For instance if you have 4 categories a a relative distibuiton of 1;10;100;1000 you would have an average of 3.97and a standard deviation close to 0.03. (3.97-1) / 0.03 is almost a z-score of 300. Are your categories well balanced? Do the averages have a wide range, between 1 and 5 or between 0 -6? With a zscore you will still have 6 categories and 5 categories, but you have altered the meaning of the mid category.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire