I have data from a survey that measures different attributes using a Likert-style scale that I treat as a single variable in my study. I have noticed that existing studies use a median but reports of those studies do not provide more details on what exactly they do.
For example:
Survey Item #1. The weather feels cold [1-5]
Survey Item #2. The weather forecast says it could be cold [1-5]
Survey Item #3. People on the streets are dressed for the cold [1-5]
How would I condense data from these three survey items into a single item that I can use to relate these to, let's say, the traffic conditions measured as High and Low using a Chi-Square analysis?
Hi,
There's a very good discussion about how to combine responses from individual questions on a likert-scale into a single factor here: https://www.researchgate.net/post/Compare_different_factors_in_questionnaire.
In general I would not use a chi-square test (assuming we're talking about Pearson's chi-square or one of its variants) with ordinal data. Although it is possible to put the information into a cross table and do a chi-square test, that would mean ignoring valuable information about the ordinal nature of your data.
Depending on how many independent variables you want to test for and what your aims are, you may want to consider a Mann-Whitney U test, ordinal regression analysis, or multi dimensional scaling instead.
If you are interested in examples of statistical tests/approaches to ordinal data, feel free to check out my paper "Student attitudes toward teaching English with Technology" where I used all of these methods (full text available from my profile).
Good luck
Are these scales self–made? Then the first step should be a proper examination of scale internal consistency and discriminant validity. There's more than one way to do it. Preferred is a confirmatory factor analysis.
Hello Nitin Reddy Katkam, My answer is really a different example but useful in terms of what you say with the high / low in that my MPH was on diabetes and according to a continuous measure of FPG or OGTT, people were classified as not having or having disease so you need some cut-point to classify. I don't know whether the weather forecast have criteria for classification - sorry. Thanks Debbie
- Badges
- Science topic