Reply With Quote
For ratio scales, use Pearson's and for ranked data (ordinal) use Spearman's rank correlation.
Hi I have a very common but somehow still confusing question regarding correct correlation to use when addressing the relationship between two liker-scale scored tests (eg- the NEO personality test and the Uses of music inventory...both measures use a likert-scale scoring system eg dont agree, somewhat agree etc, and are then converted into scores for each domain of the personality test and factors of the uses of music inventory). I have done some scatterplots for a few of the correlations between domains and factors that were expected to correlate (at least somewhat) and have found a very clumped, only marginally linear relationship.
Could ANYONE please help me regarding which correlation is better to use and why??? I am thinking spearman's given that the relationship is not strongly linear...but is this type of data ranked??? It is in a way given the use of the likert-scales...but these have been converted into scores...does that change things???
Clarification would be very appreciated.
Cheers
For ratio scales, use Pearson's and for ranked data (ordinal) use Spearman's rank correlation.
Posting Permissions
|
|