Hello,
I need to analyze data with a Spearman’s correlation test. I should find the correlation between one ordinal and one continuous variable. The ordinal variable goes from 0 to 5, where 0 represents the absence of lesions, and 5 represents presence of severe lesions.
I have been reading that to rank data for a Spearman’s correlation test I should give “rank 1” to to the highest value of the variable, “rank 2“ to the second and so on. For the continuous variable this is clear. However, I don’t understand if I also have to do it with the ordinal variable or if it is already considered ranked. If it has to be ranked as well, then does it mean that I have to give a “rank 1” to the individuals with a lesion score of 5, since it is the highest value?
I hope my doubts are clear. Thank you in advance for the answers!
Hello Marta,
If I understand your question/problem, I think it doesn't matter that your ordinal variable is the "wrong way around". You simply need to keep in mind that the correlation you obtain will be "upside down". It might simply be easier, and more logical, however, if you start with a rank of 1 to the lowest score on your ordinal variable.
You need rank the two variables of your observations in the same way: from highest to lowest or vice versa.
Robert Trevethan Babak Jamshidi Thank you for your answers!
So if I understand correctly:
- I either give rank 1 to the highest value of the continuous variable and rank 1 to score 5 of the ordinal variable (so rank 1 to the highest value in both variables),
- or give rank 1 to the lowest value of the continuous variable and to score 0 of the ordinal variable (so rank 1 to the lowest value in both). (which may seem more logical in my case)
Correct?
Hello Marta,
Given what I understand your data to be, I think that starting with giving a rank of 1 to the lowest scores on both variables would be most logical. It would also produce a correlation coefficient that was the most readily interpretable.
All the best with your research.
Dear @Robert Trevethan, the results and interpretation are the same for both scaling (ranking).
Babak Jamshidi, I suspect we're in agreement with each other. The only "issue" is that, if Marta had followed one strategy within her (original) question, she would have had to take into account that a positive correlation would have indicated an inverse relationship between her variables - which would have led to difficulties in interpretation.
I assume she's on the right track now.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size