I would like to caculate the correlation between two vectors. One vector represents the intensity of an emotion as continuous data between 0 and 100. The other vector represents the intensitiy of an emotion in 9 different steps (1 weakest, 9 strongest). I'm not sure if this is a discrete or categorical variable.

What meassure should I use in this case? Is Pearson the right way to go?


1 Answer 1


Because the scales are perhaps ordinal rather than numerical, I would suggest Spearman's correlation. [Because data are probably not normal, you should not trust any normal-based confidence intervals or tests provided by default in some statistical software packages.]

Below is a plot of 1000 fake data pairs $(x,y),$ sampled from a highly correlated bivariate distribution. In order to avoid massive over-plotting I have randomly jittered both variables. (Jittering is random uniform noise just for plotting purposes.)

R code:

cor(x, y, meth="sp")         # Spearman correlation
[1] 0.9164336

X = x + runif(1000, -.3,.3)  # uniform ...
Y = y + runif(1000, -.3,.3)  #  ... jittering
plot(X, Y, pch=20)

enter image description here


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.