Since your question is about the impact (i.e., a causal influence), mostly of things that you cannot vary in a study, some say you should not be attributing causal influence: ``causes are only those things that could, in principle, be treatments in experiment'' \citep[p.~954]{Holland1986, doi = {10.2307/2289064}}. Others seem to accept that you can. There are well developed methods on attributing causation. Books include those by Pearl, Rosenbaum, Morgan & WInship, Imbens & Rubin, Huntington-Klein, Cartwright, etc.

Hi Daniel Wright , thank you for your reply.

Would it be better to ask if there were differences in sustainability scores amongst different demographics then, rather than how they impact (I've created a coded sustainability score based on user answers so have averages for each category such as age, gender, etc). If so, please can you advise which stats would be appropriate? The sustainability score is not normally distributed... So I'm assuming I'd need a non-parametric test, however I'm a bit stuck as to whether the data counts as paired or independent. Thanks!

Hello Rachel Grunnill. No real world variable is normally distributed. See George Box's famous statement about this in the attached image. ;-)

Here are some questions for you.

1) Would it be sensible and defensible to use means and SDs when describing the sustainability score? If not, why not? And what measures of location and dispersion would you use instead?

2) Do you have respondents' actual ages, or just the age groups they belong to?

3) What is the sample size?

4) Are there other explanatory variables besides the ones you mentioned above?

5) Are you primarily interested in estimating the coefficients (with some specified degree of precision) for individual explanatory variables, or do you care mainly about how close the fitted values are to the actual values of Y?

Switched from causal hypotheses to something very different is a big change. Are you just interested in the bivariate associations between each of those demographic variables and sustainability? Also, some more detail about how you measured sustainability could be useful (as well as answers to Bruce Weaver 's questions).

Rachel Grunnill

If your dependent variable is continuous and differs from normal distribution, while you have categorical independent variables, kruskal wallis might help you understand the differences of sustainability scores among the variables.