The research question is if there is a correlation between language barriers and DASS-21 scale (dependent, continuous) I have 3 items that I am using as independent variables. The first item is language level. At first it was ordinal (beginner/intermediate/advanced) but I turned it into a dichotomous variable with 0 for beginner and 1 for fluent. I'm using Pearson r for correlation. The problem is the other two items are ordinal as a 7 point scale. Item 1 asks the participants the effect of language barriers at work and Item 2 the effect on their social life. I used spearman for correlation and found that there is a statistically significant one. I don't know what to use for prediction though. Multiple regression needs interval data so I can't use that.
1 Answer
You might have been taught that ordinal data require rank-based statistics (such as Wilcoxon or Spearman), in which case you were taught wrong. Most often, continuous summaries (like t-tests) provide powerful, accurate, and interpretable inference whereas rank-based statistics applied to the same analysis are inefficient, provide no prediction, and are extremely difficult to interpret. It shouldn't matter so much as to the number of points in the scale or the number of variables used.
-
$\begingroup$ Thank you. My advisor recommended Manova, but I'll try t-test $\endgroup$– AgapiCommented Dec 8, 2022 at 9:47