I could not find an answer to this specific question on the forum so ill make a new post. Thanks a lot for helping!
I have 5 Yes/No question, and 5 7-point likert scale items (that each corresponds to one particular Yes/No question). I would like to compute these into one overall variable measuring the same construct.
Ex:
Question 1: Have you been exposed to family members drug abuse? (Yes/No)
Question 6: How much would you say that drug abuse in your family would impact you? (7-point liker scale)
The way I was thinking of analyzing this is by coding Yes as 1 and No as 0, and then multiply the corresponding items together. So If Q1=Yes, Q2=6, The score would be 1*6=6. After doing this for all 5 items, I would combine them into one variable measuring an overall drug abuse impact.
My problem now is that the frequency of 0 is really high, resulting in highly skewed data. Could someone give some directions as to how I should perform the computation and analysis of the scale?
Again, thank you very much for your help.
Best Answer
Multiplying your Likert-type answers by binary ones would give you a variable that has 8 categories: 0 - "I have not been exposed to drug abuse in my family" and 1-7 Likert answers from Q6. The problem here is that "0" on your scale does not measure the impact of family members drug abuse on your respondent, while other categories do. Combining those two answers is like combining "Are you colorblind?" with "Do you consider this color blue? 1 - not much blue, 7 - very blue". The "0" category does not make much sens.
This kind of recoding would give you a censored data, where you do not know what the impact of drug abuse in family would be on the person without family members that use drugs.