Can the eta coefficient be used when the categorical and the quantitative variable is dependent? E.g. in the field, I made a prediction into a category (e.g. high, low or medium) of what I think a value of something will be; then I went to the lab and tested it. I now want to correlate my categorical prediction with the actual quantitative measured values. Can I use boxplots for visual illustration and the eta for the coefficient of correlation?
If you ran ANOVA, eta-square (which is the explained variance and thus the equivalent of r-square) will not tell you any more than the significant of those results.
Of course you can do box-plots, but if you want to find out which categories are statistically different from each other, you should do post-hoc comparisons from an ANOVA.
Thank you, David, for taking the time to answer - it is much appreciated. I'd still want to explain variance use eta-square, as long as it is a valid measure for what I am to trying to do.
But you are completely correct - I need to do post-hoc comparisons as well. Is there a specific one you'd advise for this experiment?
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking