Can a data set which contains both the data in nominal or ordinal scale as well as ratio scale be used for running spearman rank correlation , directly? Or any prior transformation is required to streamline those data? Whether any check to see that we can corretlate them or not in IBM SPSS is present? Is the same can be done with SAS?
The answer is yes for ordinal and ratio, but no for nominal with one exception. If the nominal are re-coded 0/1 for existence/non-existence of the condition, you can treat it as ordinal with the Spearman, but the results will default to the equivalent of a contingency coefficient.
Dear Chinmaya Nanda, the answer given by Mr. James is very right. In case of Spearman Rank correlation ranks are required to process the data, but the nominal scale indicates only identity, for example- name of a person, gender, rural /urban, literate/illetrate, country, etc. not the order or magnitude and distance in data, so Spearman's rank correlation can not be run.
Respected Sir,
Thank you for your response. Actually what I have done is I have coded gender as male=1 and female=2. In that data set I also had Education as Graduation which I coded 1 and so on upto PhD. Now I performed the spearmen rank correlation with independent variables as Gender, Education, Age (in years), service experience (in years) and no of training received as well as imparted with dependent variables as job Involvement score and satisfaction score (normalised individual score). Hence I got a negative relationship between one attribute with gender. Whether this methodology is right? What can we infer with that negatively correlated variables (say gender and involvement)?
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running