Correlation not reflecting actual similarity of values

#1
Hello,

I have 2 vector of data which correspond to rating in a test. Subjects perform the test and rate their result (from 0 to 10) and the next week they repeat the same procedure.
The values are "very similar" but correlation does no actually reflect it.

Since all values are around 8 correlation considers them as a cloud with low correlation, but considering that the range was from 0 to 10, the actual correlation value does not reflect that raters rated in a very similar way both test.

I am failing to understand how to analyze this information.
Thanks in advance
Dario

vector1 = [7 8 8 8 7 8 7];
vector2 = [6 8 9 7 6 5 7];
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
What was the value you got for the correlation statistic and respective p-value? Have you thought about looking at the kappa statistic? Also, is this the full dataset, seems like a small sample. You may also want to look at using the gamma statistic.
 
#3
Thanks for the reply.
That is actually the whole sample, there are 8 people.

Can you please comment a bit on why kappa or gamma are a better approach?

Thanks again
Dario
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
Kappa is used to evaluate inter-rater reliability

Gamma is a correlation measure for comparing two (ordinal/categorical) variables with only 1-5 internal groups.

Not sure what you are running, but I am pretty sure these data would not meet the criteria for normality or Pearson's correlation.