So, I am conducting a questionnaire to find out peoples bird feeding habits, whether they are a feeder or non-feeder of birds and their attitudes to garden bird health.
I will be running one statistical analysis test on associations between feeders and non-feeders on a variety of characteristics such as age, gender etc. Will my dependent variable be the feeder type (feeder or non-feeder) and my independent/predictor variable be the characteristics?
Secondly, I have a section on their bird feeding habits (just for feeders) with a range of questions such as when do they feed birds etc. But how do I analyse this data as I do not need to compare or find associations? Is this a pointless section?
Lastly, I have a 5 point likert scale with statements that participants have to select how much they agree or disagree with it and I will compare feeders answers to non-feeders answers. How do I go about analysing this? Do I do Chronbachs alpha to test the reliability and then another stats test to compare the two participant types?
1. When you are working with other categorical variables, you can test the difference in your two-category variable with a chi-square. When you are working with continuous variables, you can use a t-Test.
2. The descriptive data about the habits of feeders could be interesting, especially if no one has done prior work in this area.
3. Yes, use coefficient alpha to assess the reliability of your scale, and if it is adequate, combine the items into a single scale. As before, you can use a t-Test to make the comparison.
You have to perform reliability analysis, also exploratory factor analysis for the use of the instruments in your sample, to buy groups you can use student t
1. Will my dependent variable be the feeder type (feeder or non-feeder) and my independent/predictor variable be the characteristics?
Doing it the other way around would be more straighforward and the most "normal" way to frame the question. Given that your approach is quasi-experimental, the distinction between dependent and independent variable can be a bit arbitrary though, and it should be informed by the reasoning behind your study.
So, how would you phrase the question you want answer?
a) Is it: Do feeders and non-feeders differ in their characteristics?
--> do a t-test comparing the means for feeders vs. non-feeders (that is, provided your other measure is continuous. For something like "do bird feeders more often belong to one gender than another?", you can do a version of a Chi-Square test).
b) If your question is: Can I tell what type of "bird person" someone is if I know their other characteristics?"
--> you might do logistic regression with feeder vs. non-feeder as your dependent variable, and your characteristics as predictors.
(This would be an unusual way to go about it, and probably not what reviewers or supervisors would expect, but it's a completely valid way to frame it.)
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis