Good day to you, RGers!
If a researcher conducted a research study with research questions, of course, but no hypotheses to test, would it be fine to carry out inferential statistics, e.g., a correlation test, a Kruskal-Wallis test, etc., maybe to add to the significance of the study? How harmful would it be to do so if it was adjudged not fine?
Onipe Adabenege Yahaya
, whenever you use ChatGPT to generate an answer, you should cite it just as you would do with any other source.And I disagree with many of the points made by the AI, such as that inferential techniques in non-hypothesis testing research includes confidence intervals, because CI are used to assess whether an effect is zero. I also disagree about regression and GLMs because both require a separation between independent and dependent variables, which is the essence of hypothesis testing.
As for what the original question asked, a correlation could indeed be purely descriptive, but it usually tests whether a relationship is present. And Kruskal-Wallis is equivalent to a t-Test, so it requires a hypothesis to link an independent and dependent variable.
- Badges
- Science topic
- Similar topics
- Correction