Hi all,
I am doing an online survey which is mostly about people's experiences and subjective views. There are a couple of binary questions, a few Likert scales, some multiple answer ones, and mostly multiple choice. I have been reading a lot but I have no background in statistics at all. Right now I am looking into ANOVA and Regression analysis but I am not convinced either is a proper choice.
Could you give me an advice on what ways or tests of analysing the data might be the most appropriate? The analysis does not need to be very sophisticated, however, I would appreciate any suggestions. I am currently getting into SPSS, for the record.
I also know it might be an issue to test a lot of hypothesis at once, any advice on how to overcome the issue without getting into the Type S/M errors or Bonferroni? Is it just limiting as much as possible?
Thank you for your time, any help would be appreciated!
Hi Diana,
before you dive into a lot of hypothesis testing I would emphasize the importance of really getting to know your data by spending time exploring them using simple summary statistics (means, medians, variances) and graphics (boxplots, scatter diagrams etc). From your brief description it sounds as though most of your questions will not result in responses that directly conform to the assumptions of basic ANOVA or regression methods and you may be better to read up on alternatives rather than invest a lot of effort in techniques that will lead to potentially misleading answers. The good thing is that statisticians specializing the social sciences have spent a lot of time developing methods for the types of data you have collected. Here is one online resource, but there are dozens of books and many many webpages where you will find guidance. Good luck.
https://www.ilri.org/biometrics/TrainingResources/Documents/University%20of%20Reading/Guides/Guides%20on%20Analysis/ApprochAnalysis.pdf
Hi Diana,
before you dive into a lot of hypothesis testing I would emphasize the importance of really getting to know your data by spending time exploring them using simple summary statistics (means, medians, variances) and graphics (boxplots, scatter diagrams etc). From your brief description it sounds as though most of your questions will not result in responses that directly conform to the assumptions of basic ANOVA or regression methods and you may be better to read up on alternatives rather than invest a lot of effort in techniques that will lead to potentially misleading answers. The good thing is that statisticians specializing the social sciences have spent a lot of time developing methods for the types of data you have collected. Here is one online resource, but there are dozens of books and many many webpages where you will find guidance. Good luck.
https://www.ilri.org/biometrics/TrainingResources/Documents/University%20of%20Reading/Guides/Guides%20on%20Analysis/ApprochAnalysis.pdf
Hello - for survey research tools, one thing you might determine is whether the tool you are using is a validated one. Some validated tools provide instructions on how to calculate a score for either the entire instrument or subsets of it. If there isn't such a scoring mechanism, a very simple thing you can do is provide frequencies of the Yes/No fields and calculate some simple metrics such as mean and standard deviation for the items on the liker scales. Depending on the structure of your questions, you may be able to combine groups of questions on the likert scale -- be careful about negated questions in this case, as you will want to reverse the score. Then you can compare scores between different strata in your data, using t-tests or non-parametric tests depending on whether your normality assumptions hold. That would be a start.
There is a methodology in Item Response Theory that can incorporate varying numbers of rating levels within the same assessment domain. It is referred to as the Partial Credit Model (Masters, 1982) and is a form of the polytomous Rasch Model. It does require specialized software to run as well as a decent sample size. Feel free to message me if you have more specific questions.
https://link.springer.com/article/10.1007/BF02296272
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS