Hello, I am currently attempting the analysis for my MA dissertation - I am looking at inter-country differences between awareness and attitudes about dyslexia in China. Therefore, my participants have indicated which area of China they are from, and have then answered 5 questions designed to assess their level of awareness and knowledge of dyslexia, then a further 10 questions designed to assess their attitudes/beliefs. In terms of analysis, I want to ideally look at how their level of knowledge of dyslexia (first 5 q's) has an impact on their attitudes (following 10 q's), but I would also like to look at if which area of China they are from has an impact on awareness/attitudes. Is this possible? What analysis would be appropriate? And what is the best way to input this? Oh, and my data is Likert scale data. I would really appreciate any help! Thank you.
Hi, you can check whether a relationship exists between the area of residence and each one of the 15 questions by conducting a x2 test. If sig
The "textbook" method comparing the associations between two sets of interval-ish variables is a canonical correlation. These can get kind of messy with a large number of variables, so the CCA function in the R package PMA by Daniela Witten and colleagues have a version that produce a more sparse solution (i.e., it forces some of the coefficients to zero using something called the lasso). The multiple regions part may mean your analysis requires something different. Can you give more details about the number of regions and how many people you have responses from per region.
In general I worry interpreting p-values from doing 15 separate tests (i.e., one for each question) and assuming there are multiple regions this still would not isolate differences. I assume you are more interested in the patterns of results.
Hi Daniel Wright - participants could select from six possible regions of China, and there was also the option to select "prefer not to say". So 6 options for regions, +1 "prefer not to say" option. Respondents regions were as follows - prefer not to say: 3, North China: 3, North East China: 3, East China: 5, South Central China: 10, Southwest China: 1, and Northwest China: 5 (a total of 30 participants). I want to try and keep it as simple as possible, as I feel as though I'm stressing enough as it is with this, but I fear I may have made it too complex!
That sample is small for canonical correlations. The power for comparing these groups (even if you cluster them into a two groups). I think it is best to get some responses from people more experienced with small samples than I am. Hope the research goes well.
Daniel Wright yes, it's just a very small scale MA project. And I struggled for participants. But thank you for your help. :)
- Badges
- Science topic